This document discusses using wavelet domain saliency maps for secret communication in RGB images. It proposes a method to compute saliency maps using both approximation and detail coefficients from discrete wavelet transforms of the color channels. Higher numbers of secret bits would be embedded in less salient regions according to the saliency map. The saliency map approach is compared to other methods and could make steganography more secure by embedding data in less noticeable image regions.
IRJET-A Review on Implementation of High Dimension Colour Transform in Domain...IRJET Journal
This document reviews algorithms for detecting salient regions in images using high dimensional color transforms. It summarizes several existing methods that use color contrast, frequency analysis, and superpixel segmentation. A key method discussed creates a saliency map by finding the optimal linear combination of color coefficients in a high dimensional color space. This allows more accurate detection of salient objects versus methods using only RGB color. The performance of this high dimensional color transform method is improved by also utilizing relative location and color contrast between superpixels as learned features.
There exists a plethora of algorithms to perform image segmentation and there are several issues related to
execution time of these algorithms. Image Segmentation is nothing but label relabeling problem under
probability framework. To estimate the label configuration, an iterative optimization scheme is
implemented to alternately carry out the maximum a posteriori (MAP) estimation and the maximum
likelihood (ML) estimations. In this paper this technique is modified in such a way so that it performs
segmentation within stipulated time period. The extensive experiments shows that the results obtained are
comparable with existing algorithms. This algorithm performs faster execution than the existing algorithm
to give automatic segmentation without any human intervention. Its result match image edges very closer to
human perception.
Image segmentation by modified map ml estimationsijesajournal
Though numerous algorithms exist to perform image segmentation there are several issues
related to execution time of these algorithm. Image Segmentation is nothing but label relabeling
problem under probability framework. To estimate the label configuration, an iterative
optimization scheme is implemented to alternately carry out the maximum a posteriori (MAP)
estimation and the maximum likelihood (ML) estimations. In this paper this technique is
modified in such a way so that it performs segmentation within stipulated time period. The
extensive experiments shows that the results obtained are comparable with existing algorithms.
This algorithm performs faster execution than the existing algorithm to give automatic
segmentation without any human intervention. Its result match image edges very closer to
human perception.
This document discusses using super-resolution-based in-painting for object removal in images. It begins with an overview of in-painting and exemplar-based in-painting methods. It then proposes a new framework that combines exemplar-based in-painting with a single-image super-resolution method. This approach improves image quality by producing high-resolution outputs with less noise compared to exemplar-based in-painting alone. The document concludes the proposed method increases robustness for applications like satellite imaging and medical imaging by providing high quality images with damaged objects removed.
International Journal of Engineering and Science Invention (IJESI) is an international journal intended for professionals and researchers in all fields of computer science and electronics. IJESI publishes research articles and reviews within the whole field Engineering Science and Technology, new teaching methods, assessment, validation and the impact of new technologies and it will continue to provide information on the latest trends and developments in this ever-expanding subject. The publications of papers are selected through double peer reviewed to ensure originality, relevance, and readability. The articles published in our journal can be accessed online.
Implementation of High Dimension Colour Transform in Domain of Image ProcessingIRJET Journal
This document discusses implementing a high-dimensional color transform method for salient region detection in images. It aims to extract salient regions by designing a saliency map using global and local image features. The creation of the saliency map involves mapping colors from RGB space to a high-dimensional color space to find an optimal linear combination of color coefficients. This allows composing an accurate saliency map. The performance is further improved by using relative location and color contrast between superpixels as features and resolving the saliency estimation from an initial trimap using a learning-based algorithm. The method is analyzed on a dataset of training images.
Web image annotation by diffusion maps manifold learning algorithmijfcstjournal
Automatic image annotation is one of the most challenging problems in machine vision areas. The goal of this task is to predict number of keywords automatically for images captured in real data. Many methods are based on visual features in order to calculate similarities between image samples. But the computation cost of these approaches is very high. These methods require many training samples to be stored in memory. To lessen thisburden, a number of techniques have been developed to reduce the number
of features in a dataset. Manifold learning is a popular approach to nonlinear dimensionality reduction. In
this paper, we investigate Diffusion maps manifold learning method for webimage auto-annotation task.Diffusion maps
manifold learning method isused to reduce the dimension of some visual features. Extensive experiments and analysis onNUS-WIDE-LITE web image dataset with
different visual featuresshow how this manifold learning dimensionality reduction method can be applied effectively to image annotation.
INFORMATION SATURATION IN MULTISPECTRAL PIXEL LEVEL IMAGE FUSIONIJCI JOURNAL
The availability of imaging sensors operating in multiple spectral bands has led to the requirement of
image fusion algorithms that would combine the image from these sensors in an efficient way to give an
image that is more informative as well as perceptible to human eye. Multispectral image fusion is the
process of combining images from different spectral bands that are optically acquired. In this paper, we
used a pixel-level image fusion based on principal component analysis that combines satellite images of the
same scene from seven different spectral bands. The purpose of using principal component analysis
technique is that it is best method for Grayscale image fusion and gives better results. The main aim of
PCA technique is to reduce a large set of variables into a small set which still contains most of the
information that was present in the large set. The paper compares different parameters namely, entropy,
standard deviation, correlation coefficient etc. for different number of images fused from two to seven.
Finally, the paper shows that the information content in an image gets saturated after fusing four images.
IRJET-A Review on Implementation of High Dimension Colour Transform in Domain...IRJET Journal
This document reviews algorithms for detecting salient regions in images using high dimensional color transforms. It summarizes several existing methods that use color contrast, frequency analysis, and superpixel segmentation. A key method discussed creates a saliency map by finding the optimal linear combination of color coefficients in a high dimensional color space. This allows more accurate detection of salient objects versus methods using only RGB color. The performance of this high dimensional color transform method is improved by also utilizing relative location and color contrast between superpixels as learned features.
There exists a plethora of algorithms to perform image segmentation and there are several issues related to
execution time of these algorithms. Image Segmentation is nothing but label relabeling problem under
probability framework. To estimate the label configuration, an iterative optimization scheme is
implemented to alternately carry out the maximum a posteriori (MAP) estimation and the maximum
likelihood (ML) estimations. In this paper this technique is modified in such a way so that it performs
segmentation within stipulated time period. The extensive experiments shows that the results obtained are
comparable with existing algorithms. This algorithm performs faster execution than the existing algorithm
to give automatic segmentation without any human intervention. Its result match image edges very closer to
human perception.
Image segmentation by modified map ml estimationsijesajournal
Though numerous algorithms exist to perform image segmentation there are several issues
related to execution time of these algorithm. Image Segmentation is nothing but label relabeling
problem under probability framework. To estimate the label configuration, an iterative
optimization scheme is implemented to alternately carry out the maximum a posteriori (MAP)
estimation and the maximum likelihood (ML) estimations. In this paper this technique is
modified in such a way so that it performs segmentation within stipulated time period. The
extensive experiments shows that the results obtained are comparable with existing algorithms.
This algorithm performs faster execution than the existing algorithm to give automatic
segmentation without any human intervention. Its result match image edges very closer to
human perception.
This document discusses using super-resolution-based in-painting for object removal in images. It begins with an overview of in-painting and exemplar-based in-painting methods. It then proposes a new framework that combines exemplar-based in-painting with a single-image super-resolution method. This approach improves image quality by producing high-resolution outputs with less noise compared to exemplar-based in-painting alone. The document concludes the proposed method increases robustness for applications like satellite imaging and medical imaging by providing high quality images with damaged objects removed.
International Journal of Engineering and Science Invention (IJESI) is an international journal intended for professionals and researchers in all fields of computer science and electronics. IJESI publishes research articles and reviews within the whole field Engineering Science and Technology, new teaching methods, assessment, validation and the impact of new technologies and it will continue to provide information on the latest trends and developments in this ever-expanding subject. The publications of papers are selected through double peer reviewed to ensure originality, relevance, and readability. The articles published in our journal can be accessed online.
Implementation of High Dimension Colour Transform in Domain of Image ProcessingIRJET Journal
This document discusses implementing a high-dimensional color transform method for salient region detection in images. It aims to extract salient regions by designing a saliency map using global and local image features. The creation of the saliency map involves mapping colors from RGB space to a high-dimensional color space to find an optimal linear combination of color coefficients. This allows composing an accurate saliency map. The performance is further improved by using relative location and color contrast between superpixels as features and resolving the saliency estimation from an initial trimap using a learning-based algorithm. The method is analyzed on a dataset of training images.
Web image annotation by diffusion maps manifold learning algorithmijfcstjournal
Automatic image annotation is one of the most challenging problems in machine vision areas. The goal of this task is to predict number of keywords automatically for images captured in real data. Many methods are based on visual features in order to calculate similarities between image samples. But the computation cost of these approaches is very high. These methods require many training samples to be stored in memory. To lessen thisburden, a number of techniques have been developed to reduce the number
of features in a dataset. Manifold learning is a popular approach to nonlinear dimensionality reduction. In
this paper, we investigate Diffusion maps manifold learning method for webimage auto-annotation task.Diffusion maps
manifold learning method isused to reduce the dimension of some visual features. Extensive experiments and analysis onNUS-WIDE-LITE web image dataset with
different visual featuresshow how this manifold learning dimensionality reduction method can be applied effectively to image annotation.
INFORMATION SATURATION IN MULTISPECTRAL PIXEL LEVEL IMAGE FUSIONIJCI JOURNAL
The availability of imaging sensors operating in multiple spectral bands has led to the requirement of
image fusion algorithms that would combine the image from these sensors in an efficient way to give an
image that is more informative as well as perceptible to human eye. Multispectral image fusion is the
process of combining images from different spectral bands that are optically acquired. In this paper, we
used a pixel-level image fusion based on principal component analysis that combines satellite images of the
same scene from seven different spectral bands. The purpose of using principal component analysis
technique is that it is best method for Grayscale image fusion and gives better results. The main aim of
PCA technique is to reduce a large set of variables into a small set which still contains most of the
information that was present in the large set. The paper compares different parameters namely, entropy,
standard deviation, correlation coefficient etc. for different number of images fused from two to seven.
Finally, the paper shows that the information content in an image gets saturated after fusing four images.
IOSR Journal of Computer Engineering (IOSR-JCE) is a double blind peer reviewed International Journal that provides rapid publication (within a month) of articles in all areas of computer engineering and its applications. The journal welcomes publications of high quality papers on theoretical developments and practical applications in computer technology. Original research papers, state-of-the-art reviews, and high quality technical notes are invited for publications.
Computer apparition plays the most important role in human perception, which is limited to only the visual band of the electromagnetic spectrum. The need for Radar imaging systems, to recover some sources that
are not within human visual band, is raised. This paper present new algorithm for Synthetic Aperture Radar (SAR) images segmentation based on thresholding technique. Entropy based image thresholding has
received sustainable interest in recent years. It is an important concept in the area of image processing.
Pal (1996) proposed a cross entropy thresholding method based on Gaussian distribution for bi-modal images. Our method is derived from Pal method that segment images using cross entropy thresholding based on Gamma distribution and can handle bi-modal and multimodal images. Our method is tested using
Synthetic Aperture Radar (SAR) images and it gave good results for bi-modal and multimodal images. The
results obtained are encouraging.
Research Inventy : International Journal of Engineering and Science is published by the group of young academic and industrial researchers with 12 Issues per year. It is an online as well as print version open access journal that provides rapid publication (monthly) of articles in all areas of the subject such as: civil, mechanical, chemical, electronic and computer engineering as well as production and information technology. The Journal welcomes the submission of manuscripts that meet the general criteria of significance and scientific excellence. Papers will be published by rapid process within 20 days after acceptance and peer review process takes only 7 days. All articles published in Research Inventy will be peer-reviewed.
International Journal of Research in Engineering and Science is an open access peer-reviewed international forum for scientists involved in research to publish quality and refereed papers. Papers reporting original research or experimentally proved review work are welcome. Papers for publication are selected through peer review to ensure originality, relevance, and readability.
The document discusses several papers on counterfactual explanations for convolutional neural networks (CNNs). It summarizes four papers:
1. "Explaining Image Classifiers by Counterfactual Generation" which finds important image regions affecting a classifier's decision using masked and GAN-generated images.
2. "Counterfactual Visual Explanations" which generates counterfactual images to change a model's prediction by replacing image regions.
3. "Global Explanations of Convolutional Neural Networks With Concept Attribution" which measures concept importance to model predictions.
4. "Generative Counterfactual Introspection for Explainable Deep Learning" which aims to minimally change an input to alter the model's prediction.
An Experiment with Sparse Field and Localized Region Based Active Contour Int...CSCJournals
This paper discusses various experiments conducted on different types of Level Sets interactive segmentation techniques using Matlab software, on select images. The objective is to assess the effectiveness on specific natural images, which have complex image composition in terms of intensity, colour mix, indistinct object boundary, low contrast, etc. Besides visual assessment, measures such as Jaccard Index, Dice Coefficient and Hausdorrf Distance have been computed to assess the accuracy of these techniques, between segmented and ground truth images. This paper particularly discusses Sparse Field Matrix and Localized Region Based Active Contours, both based on Level Sets. These techniques were not found to be effective where object boundary is not very distinct and/or has low contrast with background. Also, the techniques were ineffective on such images where foreground object stretches up to the image boundary.
Image Denoising Based On Sparse Representation In A Probabilistic FrameworkCSCJournals
Image denoising is an interesting inverse problem. By denoising we mean finding a clean image, given a noisy one. In this paper, we propose a novel image denoising technique based on the generalized k density model as an extension to the probabilistic framework for solving image denoising problem. The approach is based on using overcomplete basis dictionary for sparsely representing the image under interest. To learn the overcomplete basis, we used the generalized k density model based ICA. The learned dictionary used after that for denoising speech signals and other images. Experimental results confirm the effectiveness of the proposed method for image denoising. The comparison with other denoising methods is also made and it is shown that the proposed method produces the best denoising effect.
This document summarizes a research paper on background subtraction techniques for motion detection in video. It describes a proposed technique that stores and compares past pixel values to the current value to determine if a pixel belongs to the background or foreground. It also discusses using a k-means algorithm and Gaussian mixture model to build a probabilistic background model and classify pixels. The paper evaluates different shadow detection approaches and finds RGB color spaces perform best for segmentation and shadow removal.
SINGLE IMAGE SUPER RESOLUTION: A COMPARATIVE STUDYcsandit
The majority of applications requiring high resolution images to derive and analyze data
accurately and easily. Image super resolution is playing an effective role in those applications.
Image super resolution is the process of producing high resolution image from low resolution
image. In this paper, we study various image super resolution techniques with respect to the
quality of results and processing time. This comparative study introduces a comparison between
four algorithms of single image super-resolution. For fair comparison, the compared algorithms
are tested on the same dataset and same platform to show the major advantages of one over the
others.
Effective Pixel Interpolation for Image Super ResolutionIOSR Journals
In the near future, there is an eminent demand for High Resolution images. In order to fulfil this
demand, Super Resolution (SR) is an approach used to renovate High Resolution (HR) image from one or more
Low Resolution (LR) images. The aspiration of SR is to dig up the self-sufficient information from each LR
image in that set and combine the information into a single HR image. Conventional interpolation methods can
produce sharp edges; however, they are approximators and tend to weaken fine structure. In order to overcome
the drawback, a new approach of Effective Pixel Interpolation method is incorporated. It has been numerically
verified that the resulting algorithm reinstate sharp edges and enhance fine structures satisfactorily,
outperforming conventional methods. The suggested algorithm has also proved efficient enough to be applicable
for real-time processing for resolution enhancement of image. Statistical examples are shown to verify the claim.
Image fusion technology is also used to fuse two processed images obtained through the algorithm
NUMBER PLATE IMAGE DETECTION FOR FAST MOTION VEHICLES USING BLUR KERNEL ESTIM...paperpublications3
This document discusses a proposed method for detecting number plates on images of fast moving vehicles that have been blurred due to motion. It begins with an introduction to image processing and digital images. It then discusses estimating the blur kernel caused by vehicle motion in order to model it as a linear uniform blur with parameters for angle and length. Existing related works on image deblurring are reviewed. The proposed system estimates the blur kernel parameters using sparse representation and Radon transform methods, allows deblurring the image, and then uses artificial neural networks to identify numbers and characters in the deblurred image. The system is evaluated on real blurred images and shown to improve license plate recognition compared to previous methods.
Robust Clustering of Eye Movement Recordings for QuantiGiuseppe Fineschi
Characterizing the location and extent of a viewer’s interest, in terms of eye movement recordings, informs a range of investigations in image and scene viewing. We present an automatic data-driven method for accomplishing this, which clusters visual point-of-regard (POR) measurements into gazes and regions-ofinterest using the mean shift procedure. Clusters produced using this method form a structured representation of viewer interest, and at the same time are replicable and not heavily influenced by noise or outliers. Thus, they are useful in answering fine-grained questions about where and how a viewer examined an image.
Efficient Image Retrieval by Multi-view Alignment Technique with Non Negative...RSIS International
The biggest challenge in today’s world is a searching of
images in a large database. For searching of an image a
technique which can be termed as Hashing is used. Already there
are many hashing techniques are present for retrieval of an
image in a large databank. The hashing technique can be done
on images by considering the high dimensional descriptor of an
image but in the existing hashing techniques single dimensional
descriptor is used from this the performance of the probability of
distribution of an image search will not achieve as expected. And
the drawback is that giving the query input in a texture format
leads to the limitations of image search that is firstly due to the
limited keyword and the second is Annotation approach by
human is ambiguous and incomplete.
To overcome from these drawbacks a new technique has been
proposed named as Multiview Alignment Hashing technique in
which it keeps the high dimensional feature descriptor data as
well probability of distribution of an images in a database. Along
with the Multiview feature descriptor another technique can be
used that is Nonnegetive Matrix Factorization (NMF). NMF is a
popular technique used in data mining in which clustering of a
data will be takes place by considering only the non- negative
matrix value.
Human Re-identification with Global and Local Siamese Convolution Neural NetworkTELKOMNIKA JOURNAL
This document proposes a global and local structure of Siamese Convolution Neural Network (SCNN) to perform human re-identification in single-shot approaches. The network extracts features from global and local parts of input images. A decision fusion technique then combines the global and local features. Experimental results on the VIPeR dataset show the proposed method achieves a normalized Area Under Curve score of 95.75% without occlusion, outperforming using local or global features alone. With occlusion, the score is 77.5%, still better than alternatives. The method performs well for re-identification including in occlusion cases by leveraging both global and local information.
This summarizes a research paper that proposes a new approach for noise estimation on images. It uses wavelet transform because of its sparse nature, then applies a Bayesian approach by imposing a Gaussian distribution on transformed pixels. It checks image quality before noise estimation using maximum likelihood decision criteria. Then designs a new bound-based estimation process using ideas from Cramer-Rao lower bound for signals in additive white Gaussian noise. The experimental results show visually better output after reconstructing the original image.
COLOUR IMAGE REPRESENTION OF MULTISPECTRAL IMAGE FUSIONacijjournal
The availability of imaging sensors operating in multiple spectral bands has led to the requirement of
image fusion algorithms that would combine the image from these sensors in an efficient way to give an
image that is more perceptible to human eye. Multispectral Image fusion is the process of combining
images optically acquired in more than one spectral band. In this paper, we present a pixel-level image
fusion that combines four images from four different spectral bands namely near infrared(0.76-0.90um),
mid infrared(1.55-1.75um),thermal- infrared(10.4-12.5um) and mid infrared(2.08-2.35um) to give a
composite colour image. The work coalesces a fusion technique that involves linear transformation based
on Cholesky decomposition of the covariance matrix of source data that converts multispectral source
images which are in grayscale into colour image. This work is composed of different segments that
includes estimation of covariance matrix of images, cholesky decomposition and transformation ones.
Finally, the fused colour image is compared with the fused image obtained by PCA transformation.
Design secure multi-level communication system based on duffing chaotic map a...IJEECSIAES
Cryptography and steganography are among the most important sciences that have been properly used to keep confidential data from potential spies and hackers. They can be used separately or together. Encryption involves the basic principle of instantaneous conversion of valuable information into a specific form that unauthorized persons will not understand to decrypt it. While steganography is the science of embedding confidential data inside a cover, in a way that cannot be recognized or seen by the human eye. This paper presents a high-resolution chaotic approach applied to images that hide information. A more secure and reliable system is designed to properly include confidential data transmitted through transmission channels. This is done by working the use of encryption and steganography together. This work proposed a new method that achieves a very high level of hidden information based on non-uniform systems by generating a random index vector (RIV) for hidden data within least significant bit (LSB) image pixels. This method prevents the reduction of image quality. The simulation results also show that the peak signal to noise ratio (PSNR) is up to 74.87 dB and the mean square error (MSE) values is up to 0.0828, which sufficiently indicates the effectiveness of the proposed algorithm.
Design secure multi-level communication system based on duffing chaotic map a...nooriasukmaningtyas
This document proposes a new method for secure multi-level communication using duffing chaotic maps and steganography. The method generates a random index vector (RIV) to determine pixel locations for embedding secret data in the least significant bits of an image. Simulation results show the proposed method achieves high quality steganography with a peak signal to noise ratio up to 74.87 dB and low mean squared error of 0.0828, indicating effectiveness. The document also reviews related work in chaotic maps, image steganography techniques, and metrics for evaluating steganography systems based on mean squared error, peak signal to noise ratio, and structural similarity index.
Semi-Supervised Method of Multiple Object Segmentation with a Region Labeling...sipij
Efficient and efficient multiple object segmentation is an important task in computer vision and object recognition. In this work; we address a method to effectively discover a user’s concept when multiple objects of interest are involved in content based image retrieval. The proposed method incorporate a framework for multiple object retrieval using semi-supervised method of similar region merging and flood fill which models the spatial and appearance relations among image pixels. To improve the effectiveness of similarity based region merging we propose a new similarity based object retrieval. The users only need to roughly indicate the after which steps desired objects contour is obtained during the automatic merging of similar regions. A novel similarity based region merging mechanism is proposed to guide the merging process with the help of mean shift technique and objects detection using region labeling and flood fill. A region R is merged with its adjacent regions Q if Q has highest similarity with Q (using Bhattacharyya descriptor) among all Q’s adjacent regions. The proposed method automatically merges the regions that are initially segmented through mean shift technique, and then effectively extracts the object contour by merging all similar regions. Extensive experiments are performed on 12 object classes (224 images total) show promising results.
1) The document discusses methods for counting people in crowded environments using computer vision techniques. It divides the main approaches into low-level analysis, foreground segmentation models, motion models, shape models, and multi-target tracking.
2) The author's approach uses a shape model based on head and shoulder detection combined with a uniform motion model from optical flow to generate probability maps for head detections. Detections are associated across frames and validated via tracking to reduce false positives and provide a person count.
3) Experimental results on standard video sequences demonstrate the method provides person counts comparable to state-of-the-art while also enabling tracking of individuals in crowded scenes.
Object Classification of Satellite Images Using Cluster Repulsion Based Kerne...IOSR Journals
Abstract: We investigated the Classification of satellite images and multispectral remote sensing data .we
focused on uncertainty analysis in the produced land-cover maps .we proposed an efficient technique for
classifying the multispectral satellite images using Support Vector Machine (SVM) into road area, building area
and green area. We carried out classification in three modules namely (a) Preprocessing using Gaussian
filtering and conversion from conversion of RGB to Lab color space image (b) object segmentation using
proposed Cluster repulsion based kernel Fuzzy C- Means (FCM) and (c) classification using one-to-many SVM
classifier. The goal of this research is to provide the efficiency in classification of satellite images using the
object-based image analysis. The proposed work is evaluated using the satellite images and the accuracy of the
proposed work is compared to FCM based classification. The results showed that the proposed technique has
achieved better results reaching an accuracy of 79%, 84%, 81% and 97.9% for road, tree, building and vehicle
classification respectively.
Keywords:-Satellite image, FCM Clustering, Classification, SVM classifier.
IOSR Journal of Computer Engineering (IOSR-JCE) is a double blind peer reviewed International Journal that provides rapid publication (within a month) of articles in all areas of computer engineering and its applications. The journal welcomes publications of high quality papers on theoretical developments and practical applications in computer technology. Original research papers, state-of-the-art reviews, and high quality technical notes are invited for publications.
Computer apparition plays the most important role in human perception, which is limited to only the visual band of the electromagnetic spectrum. The need for Radar imaging systems, to recover some sources that
are not within human visual band, is raised. This paper present new algorithm for Synthetic Aperture Radar (SAR) images segmentation based on thresholding technique. Entropy based image thresholding has
received sustainable interest in recent years. It is an important concept in the area of image processing.
Pal (1996) proposed a cross entropy thresholding method based on Gaussian distribution for bi-modal images. Our method is derived from Pal method that segment images using cross entropy thresholding based on Gamma distribution and can handle bi-modal and multimodal images. Our method is tested using
Synthetic Aperture Radar (SAR) images and it gave good results for bi-modal and multimodal images. The
results obtained are encouraging.
Research Inventy : International Journal of Engineering and Science is published by the group of young academic and industrial researchers with 12 Issues per year. It is an online as well as print version open access journal that provides rapid publication (monthly) of articles in all areas of the subject such as: civil, mechanical, chemical, electronic and computer engineering as well as production and information technology. The Journal welcomes the submission of manuscripts that meet the general criteria of significance and scientific excellence. Papers will be published by rapid process within 20 days after acceptance and peer review process takes only 7 days. All articles published in Research Inventy will be peer-reviewed.
International Journal of Research in Engineering and Science is an open access peer-reviewed international forum for scientists involved in research to publish quality and refereed papers. Papers reporting original research or experimentally proved review work are welcome. Papers for publication are selected through peer review to ensure originality, relevance, and readability.
The document discusses several papers on counterfactual explanations for convolutional neural networks (CNNs). It summarizes four papers:
1. "Explaining Image Classifiers by Counterfactual Generation" which finds important image regions affecting a classifier's decision using masked and GAN-generated images.
2. "Counterfactual Visual Explanations" which generates counterfactual images to change a model's prediction by replacing image regions.
3. "Global Explanations of Convolutional Neural Networks With Concept Attribution" which measures concept importance to model predictions.
4. "Generative Counterfactual Introspection for Explainable Deep Learning" which aims to minimally change an input to alter the model's prediction.
An Experiment with Sparse Field and Localized Region Based Active Contour Int...CSCJournals
This paper discusses various experiments conducted on different types of Level Sets interactive segmentation techniques using Matlab software, on select images. The objective is to assess the effectiveness on specific natural images, which have complex image composition in terms of intensity, colour mix, indistinct object boundary, low contrast, etc. Besides visual assessment, measures such as Jaccard Index, Dice Coefficient and Hausdorrf Distance have been computed to assess the accuracy of these techniques, between segmented and ground truth images. This paper particularly discusses Sparse Field Matrix and Localized Region Based Active Contours, both based on Level Sets. These techniques were not found to be effective where object boundary is not very distinct and/or has low contrast with background. Also, the techniques were ineffective on such images where foreground object stretches up to the image boundary.
Image Denoising Based On Sparse Representation In A Probabilistic FrameworkCSCJournals
Image denoising is an interesting inverse problem. By denoising we mean finding a clean image, given a noisy one. In this paper, we propose a novel image denoising technique based on the generalized k density model as an extension to the probabilistic framework for solving image denoising problem. The approach is based on using overcomplete basis dictionary for sparsely representing the image under interest. To learn the overcomplete basis, we used the generalized k density model based ICA. The learned dictionary used after that for denoising speech signals and other images. Experimental results confirm the effectiveness of the proposed method for image denoising. The comparison with other denoising methods is also made and it is shown that the proposed method produces the best denoising effect.
This document summarizes a research paper on background subtraction techniques for motion detection in video. It describes a proposed technique that stores and compares past pixel values to the current value to determine if a pixel belongs to the background or foreground. It also discusses using a k-means algorithm and Gaussian mixture model to build a probabilistic background model and classify pixels. The paper evaluates different shadow detection approaches and finds RGB color spaces perform best for segmentation and shadow removal.
SINGLE IMAGE SUPER RESOLUTION: A COMPARATIVE STUDYcsandit
The majority of applications requiring high resolution images to derive and analyze data
accurately and easily. Image super resolution is playing an effective role in those applications.
Image super resolution is the process of producing high resolution image from low resolution
image. In this paper, we study various image super resolution techniques with respect to the
quality of results and processing time. This comparative study introduces a comparison between
four algorithms of single image super-resolution. For fair comparison, the compared algorithms
are tested on the same dataset and same platform to show the major advantages of one over the
others.
Effective Pixel Interpolation for Image Super ResolutionIOSR Journals
In the near future, there is an eminent demand for High Resolution images. In order to fulfil this
demand, Super Resolution (SR) is an approach used to renovate High Resolution (HR) image from one or more
Low Resolution (LR) images. The aspiration of SR is to dig up the self-sufficient information from each LR
image in that set and combine the information into a single HR image. Conventional interpolation methods can
produce sharp edges; however, they are approximators and tend to weaken fine structure. In order to overcome
the drawback, a new approach of Effective Pixel Interpolation method is incorporated. It has been numerically
verified that the resulting algorithm reinstate sharp edges and enhance fine structures satisfactorily,
outperforming conventional methods. The suggested algorithm has also proved efficient enough to be applicable
for real-time processing for resolution enhancement of image. Statistical examples are shown to verify the claim.
Image fusion technology is also used to fuse two processed images obtained through the algorithm
NUMBER PLATE IMAGE DETECTION FOR FAST MOTION VEHICLES USING BLUR KERNEL ESTIM...paperpublications3
This document discusses a proposed method for detecting number plates on images of fast moving vehicles that have been blurred due to motion. It begins with an introduction to image processing and digital images. It then discusses estimating the blur kernel caused by vehicle motion in order to model it as a linear uniform blur with parameters for angle and length. Existing related works on image deblurring are reviewed. The proposed system estimates the blur kernel parameters using sparse representation and Radon transform methods, allows deblurring the image, and then uses artificial neural networks to identify numbers and characters in the deblurred image. The system is evaluated on real blurred images and shown to improve license plate recognition compared to previous methods.
Robust Clustering of Eye Movement Recordings for QuantiGiuseppe Fineschi
Characterizing the location and extent of a viewer’s interest, in terms of eye movement recordings, informs a range of investigations in image and scene viewing. We present an automatic data-driven method for accomplishing this, which clusters visual point-of-regard (POR) measurements into gazes and regions-ofinterest using the mean shift procedure. Clusters produced using this method form a structured representation of viewer interest, and at the same time are replicable and not heavily influenced by noise or outliers. Thus, they are useful in answering fine-grained questions about where and how a viewer examined an image.
Efficient Image Retrieval by Multi-view Alignment Technique with Non Negative...RSIS International
The biggest challenge in today’s world is a searching of
images in a large database. For searching of an image a
technique which can be termed as Hashing is used. Already there
are many hashing techniques are present for retrieval of an
image in a large databank. The hashing technique can be done
on images by considering the high dimensional descriptor of an
image but in the existing hashing techniques single dimensional
descriptor is used from this the performance of the probability of
distribution of an image search will not achieve as expected. And
the drawback is that giving the query input in a texture format
leads to the limitations of image search that is firstly due to the
limited keyword and the second is Annotation approach by
human is ambiguous and incomplete.
To overcome from these drawbacks a new technique has been
proposed named as Multiview Alignment Hashing technique in
which it keeps the high dimensional feature descriptor data as
well probability of distribution of an images in a database. Along
with the Multiview feature descriptor another technique can be
used that is Nonnegetive Matrix Factorization (NMF). NMF is a
popular technique used in data mining in which clustering of a
data will be takes place by considering only the non- negative
matrix value.
Human Re-identification with Global and Local Siamese Convolution Neural NetworkTELKOMNIKA JOURNAL
This document proposes a global and local structure of Siamese Convolution Neural Network (SCNN) to perform human re-identification in single-shot approaches. The network extracts features from global and local parts of input images. A decision fusion technique then combines the global and local features. Experimental results on the VIPeR dataset show the proposed method achieves a normalized Area Under Curve score of 95.75% without occlusion, outperforming using local or global features alone. With occlusion, the score is 77.5%, still better than alternatives. The method performs well for re-identification including in occlusion cases by leveraging both global and local information.
This summarizes a research paper that proposes a new approach for noise estimation on images. It uses wavelet transform because of its sparse nature, then applies a Bayesian approach by imposing a Gaussian distribution on transformed pixels. It checks image quality before noise estimation using maximum likelihood decision criteria. Then designs a new bound-based estimation process using ideas from Cramer-Rao lower bound for signals in additive white Gaussian noise. The experimental results show visually better output after reconstructing the original image.
COLOUR IMAGE REPRESENTION OF MULTISPECTRAL IMAGE FUSIONacijjournal
The availability of imaging sensors operating in multiple spectral bands has led to the requirement of
image fusion algorithms that would combine the image from these sensors in an efficient way to give an
image that is more perceptible to human eye. Multispectral Image fusion is the process of combining
images optically acquired in more than one spectral band. In this paper, we present a pixel-level image
fusion that combines four images from four different spectral bands namely near infrared(0.76-0.90um),
mid infrared(1.55-1.75um),thermal- infrared(10.4-12.5um) and mid infrared(2.08-2.35um) to give a
composite colour image. The work coalesces a fusion technique that involves linear transformation based
on Cholesky decomposition of the covariance matrix of source data that converts multispectral source
images which are in grayscale into colour image. This work is composed of different segments that
includes estimation of covariance matrix of images, cholesky decomposition and transformation ones.
Finally, the fused colour image is compared with the fused image obtained by PCA transformation.
Design secure multi-level communication system based on duffing chaotic map a...IJEECSIAES
Cryptography and steganography are among the most important sciences that have been properly used to keep confidential data from potential spies and hackers. They can be used separately or together. Encryption involves the basic principle of instantaneous conversion of valuable information into a specific form that unauthorized persons will not understand to decrypt it. While steganography is the science of embedding confidential data inside a cover, in a way that cannot be recognized or seen by the human eye. This paper presents a high-resolution chaotic approach applied to images that hide information. A more secure and reliable system is designed to properly include confidential data transmitted through transmission channels. This is done by working the use of encryption and steganography together. This work proposed a new method that achieves a very high level of hidden information based on non-uniform systems by generating a random index vector (RIV) for hidden data within least significant bit (LSB) image pixels. This method prevents the reduction of image quality. The simulation results also show that the peak signal to noise ratio (PSNR) is up to 74.87 dB and the mean square error (MSE) values is up to 0.0828, which sufficiently indicates the effectiveness of the proposed algorithm.
Design secure multi-level communication system based on duffing chaotic map a...nooriasukmaningtyas
This document proposes a new method for secure multi-level communication using duffing chaotic maps and steganography. The method generates a random index vector (RIV) to determine pixel locations for embedding secret data in the least significant bits of an image. Simulation results show the proposed method achieves high quality steganography with a peak signal to noise ratio up to 74.87 dB and low mean squared error of 0.0828, indicating effectiveness. The document also reviews related work in chaotic maps, image steganography techniques, and metrics for evaluating steganography systems based on mean squared error, peak signal to noise ratio, and structural similarity index.
Semi-Supervised Method of Multiple Object Segmentation with a Region Labeling...sipij
Efficient and efficient multiple object segmentation is an important task in computer vision and object recognition. In this work; we address a method to effectively discover a user’s concept when multiple objects of interest are involved in content based image retrieval. The proposed method incorporate a framework for multiple object retrieval using semi-supervised method of similar region merging and flood fill which models the spatial and appearance relations among image pixels. To improve the effectiveness of similarity based region merging we propose a new similarity based object retrieval. The users only need to roughly indicate the after which steps desired objects contour is obtained during the automatic merging of similar regions. A novel similarity based region merging mechanism is proposed to guide the merging process with the help of mean shift technique and objects detection using region labeling and flood fill. A region R is merged with its adjacent regions Q if Q has highest similarity with Q (using Bhattacharyya descriptor) among all Q’s adjacent regions. The proposed method automatically merges the regions that are initially segmented through mean shift technique, and then effectively extracts the object contour by merging all similar regions. Extensive experiments are performed on 12 object classes (224 images total) show promising results.
1) The document discusses methods for counting people in crowded environments using computer vision techniques. It divides the main approaches into low-level analysis, foreground segmentation models, motion models, shape models, and multi-target tracking.
2) The author's approach uses a shape model based on head and shoulder detection combined with a uniform motion model from optical flow to generate probability maps for head detections. Detections are associated across frames and validated via tracking to reduce false positives and provide a person count.
3) Experimental results on standard video sequences demonstrate the method provides person counts comparable to state-of-the-art while also enabling tracking of individuals in crowded scenes.
Object Classification of Satellite Images Using Cluster Repulsion Based Kerne...IOSR Journals
Abstract: We investigated the Classification of satellite images and multispectral remote sensing data .we
focused on uncertainty analysis in the produced land-cover maps .we proposed an efficient technique for
classifying the multispectral satellite images using Support Vector Machine (SVM) into road area, building area
and green area. We carried out classification in three modules namely (a) Preprocessing using Gaussian
filtering and conversion from conversion of RGB to Lab color space image (b) object segmentation using
proposed Cluster repulsion based kernel Fuzzy C- Means (FCM) and (c) classification using one-to-many SVM
classifier. The goal of this research is to provide the efficiency in classification of satellite images using the
object-based image analysis. The proposed work is evaluated using the satellite images and the accuracy of the
proposed work is compared to FCM based classification. The results showed that the proposed technique has
achieved better results reaching an accuracy of 79%, 84%, 81% and 97.9% for road, tree, building and vehicle
classification respectively.
Keywords:-Satellite image, FCM Clustering, Classification, SVM classifier.
This document provides an overview of salient object detection techniques, including both traditional and deep learning-based methods. It discusses early models of saliency detection based on cognitive theories of human visual attention. Global contrast and diffusion-based methods for salient object detection are described. The use of fully convolutional neural networks for deep learning-based salient object detection is also covered. Both qualitative and quantitative comparisons of detection techniques are presented. The document concludes by noting improvements in recent models from including edge and context information, but that detection remains challenging across a variety of difficult image scenarios.
The document describes a new algorithm for long-term robust visual tracking of moving objects in video sequences. The algorithm aims to overcome challenges such as geometric deformations, partial or total occlusions, and recovery after the target leaves the field of vision. It does not rely on a probabilistic process or require prior detection data. Experimental results on difficult video sequences demonstrate advantages over recent trackers. The algorithm can be used in applications like video surveillance, active vision, and industrial visual servoing.
A novel secure image steganography method based on chaos theory in spatial do...ijsptm
This paper presents a novel approach of building a secure data hiding technique in digital images. The
image steganography technique takes the advantage of limited power of human visual system (HVS). It uses
image as cover media for embedding secret message. The most important requirement for a steganographic
algorithm is to be imperceptible while maximizing the size of the payload. In this paper a method is
proposed to encrypt the secret bits of the message based on chaos theory before embedding into the cover
image. A 3-3-2 LSB insertion method has been used for image steganography. Experimental results show a
substantial improvement in the Peak Signal to Noise Ratio (PSNR) and Image Fidelity (IF) value of the
proposed technique over the base technique of 3-3-2 LSB insertion.
International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of Computer Science, Neural Networks, Electrical Engineering, Software Engineering, Information Technology, Mechanical Engineering, Chemical Engineering, Plastic Engineering, Food Technology, Textile Engineering, Nano Technology & science, Power Electronics, Electronics & Communication Engineering, Computational mathematics, Image processing, Civil Engineering, Structural Engineering, Environmental Engineering, VLSI Testing & Low Power VLSI Design etc.
Reduced-reference Video Quality Metric Using Spatial Information in Salient R...TELKOMNIKA JOURNAL
In multimedia transmission, it is important to rely on an objective quality metric which accurately
represents the subjective quality of processed images and video sequences. Maintaining acceptable
Quality of Experience in video transmission requires the ability to measure the quality of the video seen at
the receiver end. Reduced-reference metrics make use of side-information that is transmitted to the
receiver for estimating the quality of the received sequence with low complexity. This attribute enables
real-time assessment and visual degradation detection caused by transmission and compression errors. A
novel reduced-reference video quality known as the Spatial Information in Salient Regions Reduced
Reference Metric is proposed. The approach proposed makes use of spatial activity to estimate the
received sequence distortion after concealment. The statistical elements analysed in this work are based
on extracted edges and their luminance distributions. Results highlight that the proposed edge dissimilarit y
measure has a good correlation with DMOS scores from the LIVE Video Database.
A Review Of Different Approaches Of Land Cover MappingJose Katab
This document reviews different approaches for land cover mapping, including artificial neural networks (ANNs), fuzzy logic, supervised/unsupervised classification, and maximum likelihood. It discusses how each approach has been applied in previous studies for land cover classification using remote sensing data. The document also examines common problems in remote sensing image classification, such as mixed pixels, and different methods that have been proposed and used to address these issues, such as maximum likelihood classification and fuzzy classifiers. Overall, the review analyzes and compares algorithms for land cover classification and evaluates methods for overcoming problems encountered during the classification process.
This document discusses using convolutional neural networks (CNNs) to classify and segment satellite imagery. It presents a novel approach using a CNN to perform per-pixel classification of multispectral satellite imagery and a digital surface model into five categories (vegetation, ground, roads, buildings, water). The CNN is first pre-trained with unsupervised clustering then fine-tuned for classification and segmentation. Results show the CNN approach outperforms existing methods, achieving 94.49% classification accuracy and improving segmentation by reducing salt-and-pepper effects from per-pixel classification alone.
Feature Fusion and Classifier Ensemble Technique for Robust Face RecognitionCSCJournals
Face recognition is an important part of the broader biometric security systems research. In the past, researchers have explored either the Feature Space or the Classifier Space at a time to achieve efficient face recognition. In this work, both the Feature Space optimization as well as the Classifier Space optimization have been used to achieve improved results. The efficient technique of Feature Fusion in the Feature Space and Classifier Ensemble technique in the Classifier Space have been used to achieve robust and efficient face recognition. In the Feature Space, the Discrete Wavelet Transform (DWT) and the Histogram of Oriented Gradient (HOG) features have been extracted from face images and these have been used for classification purposes after Feature Fusion using the Principal Component Analysis (PCA). In the Classifier Space, a Classifier Ensemble has been used, utilizing the bagging technique for ensemble training, instead of a single classifier for efficient classification. Proper selections of various parameters of the DWT, HOG features and the Classification Ensemble have been considered to achieve optimum performance. The proposed classification technique has been applied to the AT&T (ORL) and Yale benchmark face recognition databases, and we have achieved excellent results of 99.78% and 97.72% classification accuracy respectively. The proposed Feature Fusion and Classifier Ensemble technique has been subjected to sensitivity analysis and it has been found to be robust under reduced spatial resolution conditions.
HUMAN COMPUTER INTERACTION ALGORITHM BASED ON SCENE SITUATION AWARENESScsandit
Implicit interaction based on context information is widely used and studied in the virtual scene.In context based human computer interaction, the meaning of action A is well defined. For instance, the right wave is defined turning paper or PPT in context B, And it mean volume up in context C. However, Select object in a virtual scene with multiple objects, context information is not fit. In view of this situation, this paper proposes using the least squares fitting curve beam to
predict the user's trajectory, so as to determine what object the user’s wants to operate .And fitting the starting position of the straight line according to the change of the discrete table. And
using the bounding box size control the Z variable to move in an appropriate location. Experimental results show that the proposed in this paper based on bounding box size to control
the Z variables get a good effect; by fitting the trajectory of a human hand, to predict the object that the subjects would like to operate. The correct rate is 88.6%.
Human Computer Interaction Algorithm Based on Scene Situation Awareness cscpconf
- The document proposes an algorithm for human-computer interaction based on scene situation awareness. It uses least squares fitting to predict the trajectory of a user's hand movement and determine what object the user wants to interact with.
- The algorithm first segments the user's hand from images and extracts features. It then fits the trajectory of hand movements with a nonlinear curve to predict the interaction target. Bounding box size is used to control movement in the z-axis for object selection.
- An experiment using this algorithm achieved an 88.6% correct rate in predicting the object a user intended to interact with based on fitting their hand trajectory.
Fusion of Wavelet Coefficients from Visual and Thermal Face Images for Human ...CSCJournals
In this paper we present a comparative study on fusion of visual and thermal images using different wavelet transformations. Here, coefficients of discrete wavelet transforms from both visual and thermal images are computed separately and combined. Next, inverse discrete wavelet transformation is taken in order to obtain fused face image. Both Haar and Daubechies (db2) wavelet transforms have been used to compare recognition results. For experiments IRIS Thermal/Visual Face Database was used. Experimental results using Haar and Daubechies wavelets show that the performance of the approach presented here achieves maximum success rate of 100% in many cases.
Image Retrieval using Graph based Visual SaliencyIRJET Journal
This document discusses image retrieval using graph-based visual saliency. It begins with an abstract that describes saliency detection methods and graph-based visual saliency (GBVS), which forms activation maps from image features and normalizes them to highlight salient parts. The purpose is to evaluate GBVS using statistical metrics like precision and recall, and to use genetic algorithms to improve its performance. It then provides background on saliency, different saliency approaches, what graph-based visual saliency is, its advantages and applications. Finally, it reviews several related works on visual saliency models.
A hybrid approach for analysis of dynamic changes in spatial dataijdms
Any geographic location undergoes changes over a period of time. These changes can be observed by
naked eye, only if they are huge in number spread over a small area. However, when the changes are small
and spread over a large area, it is very difficult to observe or extract the changes. Presently, there are few
methods available for tackling these types of problems, such as GRID, DBSCAN etc. However, these
existing mechanisms are not adequate for finding an accurate changes or observation which is essential
with respect to most important geometrical changes such as deforestations and land grabbing etc.,. This
paper proposes new mechanism to solve the above problem. In this proposed method, spatial image
changes are compared over a period of time taken by the satellite. Partitioning the satellite image in to
grids, employed in the proposed hybrid method, provides finer details of the image which are responsible
for improving the precision of clustering compared to whole image manipulation, used in DBSCAN, at a
time .The simplicity of DBSCAN explored while processing portioned grid portion.
This document proposes a convolutional neural network (CNN) to automatically classify aerial and remote sensing images. The CNN has six layers - three convolutional layers to extract visual features from the images at different levels of abstraction, two fully-connected layers to integrate the extracted features, and a final softmax classifier layer to classify the images. The CNN is evaluated on two datasets and is shown to outperform state-of-the-art baselines in terms of classification accuracy, demonstrating its ability to learn spatial features directly from images without relying on handcrafted features or descriptors.
This document summarizes a research paper that proposes using an artificial neural network tuned by a simulated annealing algorithm for real-time credit card fraud detection. The paper describes how simulated annealing can be used to train the weights of a neural network model to classify credit card transactions as fraudulent or non-fraudulent based on attributes of past transactions. The algorithm is tested on a real-world credit card transaction dataset and is found to effectively classify most transactions correctly, though some misclassifications still occur.
Wireless sensor networks (WSN) have been widely used in various applications.
In these networks nodes collect data from the attached sensors and send their data to a base
station. However, nodes in WSN have limited power supply in form of battery so the nodes
are expected to minimize energy consumption in order to maximize the lifetime of WSN. A
number of techniques have been proposed in the literature to reduce the energy
consumption significantly. In this paper, we propose a new clustering based technique
which is a modification of the popular LEACH algorithm. In this technique, first cluster
heads are elected using the improved LEACH algorithm as usual, and then a cluster of
nodes is formed based on the distance between node and cluster head. Finally, data from
node is transferred to cluster head. Cluster heads forward data, after applying aggregation,
to the cluster head that is closer to it than sink in forward direction or directly to the sink.
This reduction in distance travelled improves the performance over LEACH algorithm
significantly.
This document provides an overview of vertical handover decision strategies in heterogeneous wireless networks. It begins with an introduction to always best connectivity requirements in next generation networks that allow users to move between different network technologies. It then discusses the key aspects of handover management, including the three phases of initiation, decision, and execution. Various criteria for the handover decision process are described, such as received signal strength, network connection time, available bandwidth, power consumption, cost, security, and user preferences. Different types of handover decision strategies are categorized, including those based on network conditions, user preferences, multiple attributes, fuzzy logic/neural networks, and context awareness. The strategies are analyzed and their advantages/disadvantages compared.
This paper presents the design and performance comparison of a two stage
operational amplifier topology using CMOS and BiCMOS technology. This conventional op
amp circuit was designed by using RF model of BSIM3V3 in 0.6 μm CMOS technology and
0.35 μm BiCMOS technology. Both the op amp circuits were designed and simulated,
analyzed and performance parameters are compared. The performance parameters such as
gain, phase margin, CMRR, PSRR, power consumption etc achieved are compared. Finally,
we conclude the suitability of CMOS technology over BiCMOS technology for low power
RF design.
In Cognitive Radio Networks (CRN), Cooperative Spectrum Sensing (CSS) is
used to improve performance of spectrum sensing techniques used for detection of licensed
(Primary) user’s signal. In CSS, the spectrum sensing information from multiple unlicensed
(Secondary) users are combined to take final decision about presence of primary signal. The
mixing techniques used to generate final decision about presence of PU’s signal are also
called as Fusion techniques / rules. The fusion techniques are further classified as data
fusion and decision fusion techniques. In data fusion technique all the secondary users
(SUs) share their raw information of spectrum detection like detected energy or other
statistical information, while in decision fusion technique all the SUs take their local
decisions and share the decision by sending ‘0’ or ‘1’ corresponding to absence and presence
of PU’s signal respectively. The rules used in decision fusion techniques are OR rule, AND
rule and K-out-of-N rule. The CSS is further classified as distributed CSS and centralized
CSS. In distributed CSS all the SUs share the spectrum detection information with each
other and by mixing the shared information; all the SUs take final decision individually. In
centralized CSS all the SUs send their detected information to a secondary base station /
central unit which combines the shared information and takes final decision. The secondary
base station shares the final decision with all the SUs in the CRN. This paper covers
overview of information fusion methods used for CSS and analysis of decision fusion rules
with simulation results.
This paper analyzes the impact of network scalability on various physical attributes of Zigbee networks. Simulations were conducted using Qualnet to evaluate the performance of the Zigbee physical layer based on energy consumption and throughput. Energy consumption was analyzed for different modulation schemes (ASK, BPSK, OQPSK), network sizes (2-50 nodes), and clear channel assessment modes. The results showed that OQPSK and ASK had lower energy consumption than BPSK. Throughput was highest for OQPSK. While carrier sense had slightly higher throughput than other CCA modes, the energy consumption differences between CCA modes were minor.
This paper gives a brief idea of the moving objects tracking and its application.
In sport it is challenging to track and detect motion of players in video frames. Task
represents optical flow analysis to do motion detection and particle filter to track players
and taking consideration of regions with movement of players in sports video. Optical flow
vector calculation gives motion of players in video frame. This paper presents improved
Luacs Kanade algorithm explained for optical flow computation for large displacement and
more accuracy in motion estimation.
A rapid progress is seen in the field of robotics both in educational and industrial
automation sectors. The Robotics education in particular is gaining technological advances
and providing more learning opportunities. In automotive sector, there is a necessity and
demand to automate daily human activities by robot. With such an advancement and
demand for robotics, the realization of a popular computer game will help students to learn
and acquire skills in the field of robotics. The computer game such as Pacman offers
challenges on both software and hardware fronts. In software, it provides challenges in
developing algorithms for a robot to escape from the pool of attacking robots and to develop
algorithms for multiple ghost robots to attack the Pacman. On the hardware front, it
provides a challenge to integrate various systems to realize the game. This project aims to
demonstrate the pacman game in real world as well as in simulation. For simulation
purpose Player/Stage is used to develop single-client and multi-client architectures. The
multi- client architecture in player/stage uses one global simulation proxy to which all the
robot models are connected. This reduces the overhead to manage multiple robots proxy.
The single-client architecture enables only two robot models to connect to the simulation
proxy. Multi-client approach offers flexibility to add sensors to each port which will be used
distinctly by the client attached to the respective robot. The robots are named as Pacman
and Ghosts, which try to escape and attack respectively. Use of Network Camera has been
done to detect the global positions of the robots and data is shared through inter-process
communication.
In Content-Based Image Retrieval (CBIR) systems, the visual contents of the
images in the database are took out and represented by multi-dimensional characteristic
vectors. A well known CBIR system that retrieves images by unsupervised method known
as cluster based image retrieval system. For enhancing the performance and retrieval rate
of CBIR system, we fuse the visual contents of an image. Recently, we developed two
cluster-based CBIR systems by fusing the scores of two visual contents of an image. In this
paper, we analyzed the performance of the two recommended CBIR systems at different
levels of precision using images of varying sizes and resolutions. We also compared the
performance of the recommended systems with that of the other two existing CBIR systems
namely UFM and CLUE. Experimentally, we find that the recommended systems
outperform the other two existing systems and one recommended system also comparatively
performed better in every resolution of image.
Information Systems and Networks are subjected to electronic attacks. When
network attacks hit, organizations are thrown into crisis mode. From the IT department to
call centers, to the board room and beyond, all are fraught with danger until the situation is
under control. Traditional methods which are used to overcome these threats (e.g. firewall,
antivirus software, password protection etc.) do not provide complete security to the system.
This encourages the researchers to develop an Intrusion Detection System which is capable
of detecting and responding to such events. This review paper presents a comprehensive
study of Genetic Algorithm (GA) based Intrusion Detection System (IDS). It provides a
brief overview of rule-based IDS, elaborates the implementation issues of Genetic Algorithm
and also presents a comparative analysis of existing studies.
Step by step operations by which we make a group of objects in which attributes
of all the objects are nearly similar, known as clustering. So, a cluster is a collection of
objects that acquire nearly same attribute values. The property of an object in a cluster is
similar to other objects in same cluster but different with objects of other clusters.
Clustering is used in wide range of applications like pattern recognition, image processing,
data analysis, machine learning etc. Nowadays, more attention has been put on categorical
data rather than numerical data. Where, the range of numerical attributes organizes in a
class like small, medium, high, and so on. There is wide range of algorithm that used to
make clusters of given categorical data. Our approach is to enhance the working on well-
known clustering algorithm k-modes to improve accuracy of algorithm. We proposed a new
approach named “High Accuracy Clustering Algorithm for Categorical datasets”.
Brain tumor is a malformed growth of cells within brain which may be
cancerous or non-cancerous. The term ‘malformed’ indicates the existence of tumor. The
tumor may be benign or malignant and it needs medical support for further classification.
Brain tumor must be detected, diagnosed and evaluated in earliest stage. The medical
problems become grave if tumor is detected at the later stage. Out of various technologies
available for diagnosis of brain tumor, MRI is the preferred technology which enables the
diagnosis and evaluation of brain tumor. The current work presents various clustering
techniques that are employed to detect brain tumor. The classification involves classification
of images into normal and malformed (if detected the tumor). The algorithm deals with
steps such as preprocessing, segmentation, feature extraction and classification of MR brain
images. Finally, the confirmatory step is specifying the tumor area by technique called
region of interest.
A Proxy signature scheme enables a proxy signer to sign a message on behalf of
the original signer. In this paper, we propose ECDLP based solution for chen et. al [1]
scheme. We describe efficient and secure Proxy multi signature scheme that satisfy all the
proxy requirements and require only elliptic curve multiplication and elliptic curve addition
which needs less computation overhead compared to modular exponentiations also our
scheme is withstand against original signer forgery and public key substitution attack.
This document proposes a digital watermarking technique using LSB replacement with secret key insertion for enhanced data security. The technique works by inserting a watermark into the least significant bits of pixels in an image. A secret key is also inserted during transmission for additional security. The watermarked image is generated without noticeably impacting image quality. The proposed method was tested on sample images and successfully embedded watermarks while maintaining visual quality. The technique aims to provide copyright protection and authentication of digital images and documents.
Today among various medium of data transmission or storage our sensitive data
are not secured with a third-party, that we used to take help of. Cryptography plays an
important role in securing our data from malicious attack. This paper present a partial
image encryption based on bit-planes permutation using Peter De Jong chaotic map for
secure image transmission and storage. The proposed partial image encryption is a raw data
encryption method where bits of some bit-planes are shuffled among other bit-planes based
on chaotic maps proposed by Peter De Jong. By using the chaotic behavior of the Peter De
Jong map the position of all the bit-planes are permuted. The result of the several
experimental, correlation analysis and sensitivity test shows that the proposed image
encryption scheme provides an efficient and secure way for real-time image encryption and
decryption.
This paper presents a survey of Dependency Analysis of Service Oriented
Architecture (SOA) based systems. SOA presents newer aspects of dependency analysis due
to its different architectural style and programming paradigm. This paper surveys the
previous work taken on dependency analysis of service oriented systems. This study shows
the strengths and weaknesses of current approaches and tools available for dependency
analysis task in context of SOA. The main motivation of this work is to summarize the
recent approaches in this field of research, identify major issue and challenges in
dependency analysis of SOA based systems and motivate further research on this topic.
In this paper, proposed a novel implementation of a Soft-Core system using
micro-blaze processor with virtex-5 FPGA. Till now Hard-Core processors are used in
FPGA processor cores. Hard cores are a fixed gate-level IP functions within the FPGA
fabrics. Now the proposed processor is Soft-Core Processor, this is a microprocessor fully
described in software, usually in an HDL. This can be implemented by using EDK tool. In
this paper, developed a system which is having a micro-blaze processor is the combination
of both hardware & Software. By using this system, user can control and communicate all
the peripherals which are in the supported board by using Xilinx platform to develop an
embedded system. Implementing of Soft-Core process system with different peripherals like
UART interface, SPA flash interface, SRAM interface has to be designed using Xilinx
Embedded Development Kit (EDK) tools.
The article presents a simple algorithm to construct minimum spanning tree and
to find shortest path between pair of vertices in a graph. Our illustration includes the proof
of termination. The complexity analysis and simulation results have also been included.
Wimax technology has reshaped the framework of broadband wireless internet
service. It provides the internet service to unconnected or detached areas such as east South
Africa, rural areas of America and Asia region. Full duplex helpers employed with one of
the relay stations selection and indexing method that is Randomized Distributed Space Time
are used to expand the coverage area of primary Wimax station. The basic problem was
identified at cell edge due to weather conditions (rain, fog), insertion of destruction because
of multiple paths in the same communication channel and due to interference created by
other users in that communication. It is impractical task for the receiver station to decode
the transmitted signal successfully at the cell edges, which increases the high packet loss and
retransmissions. But Wimax is a outstanding technology which is used for improving the
quality of internet service and also it offers various services like Voice over Internet
Protocol, Video conferencing and Multimedia broadcast etc where a little delay in packet
transmission can cause a big loss in the communication. Even setup and initialization of
another Wimax station nearer to each other is not a good alternate, where any mobile
station can easily handover to another base station if it gets a strong signal from other one.
But in rural areas, for few numbers of customers, installation of base station nearer to each
other is costlier task. In this review article, we present a scheme using R-DSTC technique to
choose and select helpers (relay nodes) randomly to expand the coverage area and help to
mobile station as a helper to provide secure communication with base station. In this work,
we use full duplex helpers for better utilization of bandwidth.
Radio Frequency identification (RFID) technology has become emerging
technique for tracking and items identification. Depend upon the function; various RFID
technologies could be used. Drawback of passive RFID technology, associated to the range
of reading tags and assurance in difficult environmental condition, puts boundaries on
performance in the real life situation [1]. To improve the range of reading tags and
assurance, we consider implementing active backscattering tag technology. For making
mobiles of multiple radio standards in 4G network; the Software Defined Radio (SDR)
technology is used. Restrictions in Existing RFID technologies and SDR technology, can be
eliminated by the development and implementation of the Software Defined Radio (SDR)
active backscattering tag compatible with the EPC global UHF Class 1 Generation 2 (Gen2)
RFID standard. Such technology can be used for many of applications and services.
2. 440
predict which regions in image are more likely to attract human attention and to be gazed at. With the
development of brain and human vision science, progress has been made in understanding visual selective
attention in a plausible biological way, and several computational attention models have been proposed [2]-
[4]. In these models, low-level features such as orientation, intensity, motion, etc. are first extracted, and then
through non-linear biologically inspired combination of these features, an attention map (usually called
saliency map) can be generated. In this map, the interesting locations are highlighted and the intensity value
of the map represents the attention importance. Under the guidance of the attention map, resource can be
allocated non-uniformly to improve the subjective quality [5]-[8]. Although such research shows promising
results, it is still not a completely resolved problem. The use of the saliency map can be seen in object
segmentation [9, 10, 12]; visual search in complex scenes [13]; traffic signs detection [14]; image retrieval
[15]; image watermarking [16]; image compression [17]; image fusion [18]; and many other image/visual
applications. Steganography or data hiding in images may emerge as one of the major applications of
saliency map. We are working upon this application of saliency map so as to make steganography more
secure.
B. Steganography
The term steganography is derived from the Greek word steganos, meaning “covered,” and graphein, “to
write” [19]. In steganography, a message is hidden within another seemingly ordinary (cover) message in
such a way that only the sender and intended recipient will even know the hidden message exists. This idea
is different from that of cryptography, which only tries to make the message unreadable. Modern
steganography’s goal is to keep its mere presence undetectable, but steganography systems because of their
invasive nature—leave behind detectable traces in the cover medium. Due to this problem it is mandatory to
understand the statistics of Images and to make the data hidden in such a manner so that no detectable traces
are left behind.
II. LITERATURE REVIEW
A. Previous work in Saliency Map Computation
Saliency has also been referred to as visual attention [20], unpredictability, rarity, or surprise [22, 23].
Saliency estimation methods can broadly be classified as biologically based, purely computational, or a
combination. In the biological category one of the reputable works is by Itti and Koch [21] whose method
was based upon biologically plausible architecture proposed by Koch and Ullman [25]. They determine
center-surround contrast using a Difference of Gaussians (DoG) approach. Frintrop et al. [24] present a
method inspired by Itti’s method, but they compute center surround differences with square filters and use
integral images to speed up the calculations.
In computational Visual Saliency models, low-level features and the contrast approach are still used but the
model is not constructed based on any biological mechanism. The contrast is mainly obtained through the use
of Euclidian distance in different sized window filters [20, 26, 27]. The contrast images obtained in the works
of [20, 26, 27] are summed to form the final saliency map.
The third category of methods is those that incorporate ideas that are partly based on biological models and
partly on computational ones. For instance, Harel et al. [28] create feature maps using Itti’s method but
perform their normalization using a graph based approach.
Recently, there is a trend to model VS computationally in the frequency domain [11, 29]. In the works of
Hou and Zhang [11], a spectral residual approach was used to generate the saliency map. The saliency map is
the inverse of the spectral residual. The saliency map of this method is rather accurate in providing the
locations of important regions in a given visual scene but is terribly low in resolution. In [29], Achanta et al.
debated that the saliency map should have well-defined borders, uniformly highlighting the object if it is
salient, and most of all; the saliency map should be in high resolution. In the authors' opinion, without
conforming to the points mentioned [29], the saliency would have limited usefulness in certain applications.
Therefore, Achanta et al. proposed a method which generates the saliency map solely by contrast
representation. Although the approach used by Achanta et al. gives high resolution maps which has its
usefulness in some applications but in many other applications such as content based image retrieval (CBIR)
all that matters is the detection of salient objection with acceptable resolution. In fact, the approach used in
[29] will eliminate many small detailed objects and textures which could be of importance when the
smoothing is applied to the spatial domain. Furthermore, as long as the saliency map provides the correct
location of important objects and is of reasonable resolution (object can be visually identified), the map can
3. 441
be considered acceptable. In another approach Discrete wavelet transform was used by Christopher et al. [30]
to compute the saliency map. This approach is successful in detecting salient regions in an image with
acceptable resolution but it considers only the contrast of LL band, it totally ignores the other three detail
bands. This act will eliminate the smaller and finer details. But sometimes some important information may
be present in these detail components. To solve this problem we have drawn an idea about saliency model
based on wavelet transform domain by processing approximation (LL) as well as all the three detail
coefficients (LH, HL, HH) in our previous work [41]. Here we will compare our saliency map with other
techniques on the parameters of Precision, Recall and F-measure and will use the same saliency map for
steganography application. It will also be shown that saliency map makes data hiding mire secure with the
help of parameters like PSNR, MSE and Quality.
B. Previous work in Steganography
Various methods of steganography have been proposed in the literature. All these methods are broadly
classified into 3 categories. Firstly methods which are hiding data in spatial domain. One of the commonly
used techniques in this category is the LSB where the least significant bit of each pixel is replaced by bits of
the secret till secret message finishes [2, 4, 5, 6]. The risk of information being uncovered with this method
as is very much prone to ‘sequential scanning’ based techniques [1], which are threatening its security. The
random pixel manipulation technique attempts at overcoming this problem, where pixels, which will be used
to hide data are chosen in a random fashion based on a stego key. However, this key should be shared
between the entities of communication as a secret key. Moreover, some synchronization between the entities
is required when changing the key [1]. This will put key management overhead on the system. In second
category data is hidden in frequency domain. . It comprises algorithms based upon discrete cosine transforms
(DCT), Fourier transforms (FT), and discrete wavelet transforms (DWT). In this domain various algorithms
are Li and Wang steganography [31], McKeon 2DdiscreteFourier transform (DFT) based steganography [38],
Jsteg for jpeg images [32, 33], OutGuess [34], “F5’’algorithm [35]. Besides some DWT based techniques
like W.Y.Chen, Color image steganography scheme using set partitioning in hierarchical trees coding [36],
Abdulaziz and Pang technique based upon vector quantization called Linde-Buzo-Gray (LBG)coupled with
block codes known as BCH code and 1-stage discrete Haar wavelet transforms [37] are certain wavelet based
steganography techniques. The DWT-based embedding technique is still in its infancy.
The third category of steganography is Model-based steganography which introduces a different
methodology, where the message is embedded in the cover according to a model representing cover message
statistics. The model-based technique, proposed by Sallee, tries to model statistical properties of an image
and preserves them during embedding process [39]. Hioki [38], presented an adaptive method termed ‘‘A
Block Complexity based Data Embedding’’ (ABCDE). ABCDE works in a very similar method as BPCS,
but employs a more sophisticated complexity metric. This model based steganography inspires for a model
based upon human visual system which may be further be used for secure steganography.
III. SALIENCY MODEL COMPUTATION
The various calculations related to computation of saliency model are as follows:
The input colored image is first converted to Lab color space so as to make it device independent. Also the L
component distinguishes the intensity or lightness component from the color information. Then taking the all
the L, a, b components individually we performed the single level DWT decomposition. After this we got the
four individual components named LL1, HL1, LH1, HH1 for all the three L, a, b images. Then for each
component individually we calculate the contrast image using the Euclidean distance with the help of
following formula:
C(x,y) =
2
)),(( μμ lyxl − ……{Equation 1}
Where μl is the mean of LL component of L image (from L, a, b component images) and μl (x, y) is the
intensity of individual pixel of LL component of L sub-image. Similarly 12 contrast images will be
calculated for four sub-bands of each component image of Lab color space. The Euclidean distance is the
best representative of the difference in image or pixel that is why Euclidean distance has been used in this
work. Then inverse DWT operation will be performed taking four processed sub-bands of each sub-image to
get processed L, a, b sub-image Lp, ap, bp. then these processed components will be normalized to the range
[0,255].These processed sub-images will be combined to get the saliency map using the following formula:
4. 442
Sm = Lp(x,y) + ap (x,y) + bp(x,y)……{Equation 2}
where Sm is the saliency map and Lp ap, bp are processed sub-images of Lab Color space.
Again the saliency map will be renormalized to the range [0,255] to get the final saliency map. Finally we
can equalize the histogram of saliency map to get sharper saliency maps. The use of LH, HL and HH
component in addition to LL component will help a lot to determine the small objects in the images as these
small objects may be of great importance when application like steganography is being considered. As these
components (LH, HL, HH) corresponds to detail coefficients of wavelet domain which in turn represents the
fine details/small objects in horizontal, vertical and diagonal direction so their consideration while processing
for saliency map is very important.
This saliency map computation has been applied to various images of different dimensions. It has been
observed that our method generates the saliency map which is sufficient enough to distinguish and recognize
the salient objects in an image. The boundaries of the salient objects are also clearly visible.
IV. SALIENCY MAP IN STEGANOGRAPHY
Now this saliency map has been used as model for hiding secret bits of data in the image. Based on the
saliency map higher number of bits may be assigned to the points having lesser values in saliency map and
lesser number of bits to the points having higher values in saliency map.
The strategy for this division may vary from one work to another depending upon the requirements of
application. For implementation purpose we have divided the values [0-255] in eight equivalence classes
assuming maximum 3 bits will be inserted in a channel and maximum of two channels will be used in a pixel.
So we will insert 0-6 bits (7 classes) in a pixel depending upon the value of saliency map. We are not
inserting more than three bits in a channel because it has been noticed that inserting four or more bits may
cause some visible changes in statistics of image. Also higher values are more sensitive to change in color so
we are inserting zero bits of secret data in pixels having values in the range [192-255] (2 classes
corresponding to 0 bits). The concept could be better understood with the help of table 1 below.
TABLE I. STRUCTURE OF EQUIVALENCE TABLE REPRESENTING RANGE OF EQUIVALENCE CLASSES & CORRESPONDING NUMBER OF
SECRET BITS TO BE INSERTED
Sr. no Range of
Equivalence
Class
No of secret
bits to be
inserted
1. 0 to < 32 6
2. 32 to < 64 5
3. 64 to < 96 4
4. 96 to < 128 3
5. 128 to < 160 2
6. 160 to < 192 1
7. 192 to < 255 0
V. RESULTS, COMPARISON AND DISCUSSION
We have implemented the above methodology in Matlab 7.0. We have used more than 20 sets of images of
various sizes and resolution for experiment purposes.
A. Results for Saliency Map Computation
Some of the results for computing saliency map with proposed approach are shown in Figure 1. below:
5. 443
Figure. 1 Column (a) and (c) are original images and column (b) and (d) are their corresponding saliency maps
B. Comparison of Proposed Saliency Map with Other Methods
We have also compared the visual results of our image with various state of art techniques like spectral
residual approach [11], frequency tuned approach [29] and DWT based technique by Christopher et. al. [30].
The results are shown below in figure 2:
Figure 2. Implementation results: (a) original images; (b) saliency maps from spectral residual approach [11]; (c) saliency maps from
frequency-tuned approach [29]; (d) saliency maps from the explored approach - using the DWT transformed domain [30].; (e) proposed
method
It may be noticed from figure 2 above that proposed technique are not as much visually appealing or salient
as much frequency tuned approach’s results seems. But it should be understood that proposed technique is
not using any filter for smoothing so no small objects would be left out which is not the case with frequency
tuned approach [29]. Also our approach is using detail coefficients of DWT domain in addition to
approximation coefficients which is not the case with Christopher et. al. approach [30]. This will definitely
consider even small details which may be of importance with respect to human visual system. The extra cost
of processing the detail coefficients in addition to approximation coefficients can be justified in terms of
goodness of accuracy of saliency map. Precision, Recall and F-measure are various parameters which may be
used for the purpose. We have annotated a rectangle from 20 different subjects about the salient region of an
image. The axis points of rectangle in image are more or less same so a standard rectangle is taken and used
as ground truth based upon the feedback of 20 subjects.
6. 444
∑∑ ×
x y
yxsyxt )),(),((
Precision = ________________________ -------------------- {Equation 3}
∑∑x y
yxs ),(
∑∑ ×
x y
yxsyxt )),(),((
Recall = ________________________ -------------------- {Equation 4}
∑∑x y
yxt ),(
(1+σ) X P X R
F- measure = _______________ --------------------{Equation 5}
(σ X P) + R
where s(x, y) is the saliency map image and t(x, y) is a ground truth image within user annotated rectangle.
Here precision is calculated as ratio of total saliency, i.e sum of intensities in the saliency map captured
inside the user annotated rectangle to the total saliency computed for the image. Recall is calculated as the
ratio of the total saliency captured inside the user annotated rectangle to the area of rectangle. F-measure is
the overall performance measurement as the weighted harmonic mean between the precision and recall
values. σ is real and positive constant which decides the importance of precision over recall. In our work σ is
taken as 0.3 because precision is more important than recall. Precision has been taken with importance of
70% and recall with importance of 30%.
TABLE II: COMPARISON OF PROPOSED SALIENCY MAP WITH FREQUENCY TUNED [29] AND WAVELET DOMAIN BASED TECHNIQUE [30] FOR
PRECISION, RECALL AND F-MEASURE
Sr. No Image Size Technique Precision Recall F-measure
1 Star 120 X 80
Proposed 0.9726 0.5627 0.8326
Waveiee 0.9579 0.446 0.7573
Frequency_tuned 0.9566 0.1904 0.4960
2 Bridge 80 X 120
Proposed 0.7503 0.5869 0.7050
Waveiee 0.7359 0.4117 0.6227
Frequency_tuned 0.6201 0.1176 0.3122
3 Fireman 120 X 80
Proposed 0.8728 0.5606 0.7734
Waveiee 0.8529 0.3748 0.6589
Frequency_tuned 0.8057 0.2083 0.4848
4 Roses 80 X 120
Proposed 0.7613 0.6087 0.7197
Waveiee 0.7387 0.4571 0.6468
Frequency_tuned 0.7184 0.0951 0.2859
5 Scene 120 X 80
Proposed 0.5612 0.6853 0.5857
Waveiee 0.5462 0.5051 0.5361
Frequency_tuned 0.5265 0.2911 0.4437
6 Elephants 120 X 80
Proposed 0.8048 0.6947 0.7764
Waveiee 0.7996 0.4029 0.6516
Frequency_tuned 0.7708 0.1494 0.3933
7 Horses 120 X 80
Proposed 0.7593 0.5971 0.7145
Waveiee 0.746 0.5079 0.6732
Frequency_tuned 0.7298 0.1075 0.3124
8 Red_Flowers 120 X 80
Proposed 0.8043 0.5961 0.7443
Waveiee 0.7797 0.4804 0.6817
Frequency_tuned 0.7379 0.0777 0.2492
7. 445
It can be seen from the table 2. above that our technique is better in terms of precision, Recall and F-measure
for almost all the images. We have performed the experiment for 20 images approximately but due to lack of
space we have shown data only for 8 images. The details of all the images can be seen through the
comparison graphs given below for all three parameters.
Figure 3. Comparison of Precision values of Images for Proposed saliency map Frequency tuned [29] and wavelet domain based
technique [30]
Figure 3. above shows that precision value of our proposed technique stands above the precision value of
frequency tuned [29] and wavelet based technique [30] shown by waveieee in graph for almost all the
images.
Figure 4. Comparison of Recall values of Images for Proposed saliency map Frequency tuned [29] and wavelet domain based technique
[30]
Figure 4. and Figure 5. above shows that Recall value and F-measure value of our proposed technique has
clearly outshined the frequency tuned [29] and wavelet domain based technique [30] represented by waveieee
in the graphs.
C. Results for Steganography Using Proposed Saliency Map
The visual results after hiding secret information for some of the images are shown below in figure 6:
0
0.2
0.4
0.6
0.8
1
1.2
Precision_values
IMAGES
Precision
Proposed
Waveieee
Frequency_tuned
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Recall_value
IMAGES
Recall
waveieee
proposed
Frequency_tuned
8. 446
Figure 5. Comparison of F-measure values of Images for Proposed saliency map Frequency tuned [29] and wavelet domain based
technique [30]
Cover Image Stego Image Cover Image Stego Image
Cover Image Stego Image Cover Image Stego Image
(a) (b) (c) (d)
Figure: 6 Column (a) and (c) are original cover images and column (b) and (d) are their corresponding stego images
It may be analyzed that there is no much visual difference in original and stego image using the proposed
method. We have calculated the MSE and PSNR values for the experiment images with the help of following
formulas:
MSE =
NM
nmInmI
NM
*
)],(2),(1[
,
2
∑ −
In the previous equation, M and N are the number of rows and columns in the input images, respectively.
Then the block computes the PSNR using the following equation:
PSNR = 1010 Log
⎟⎟
⎠
⎞
⎜⎜
⎝
⎛
MSE
R 2
In the previous equation, R is the maximum fluctuation in the input image data type. For example, if the input
image has a double-precision floating-point data type, then R is 1. If it has an 8-bit unsigned integer data
type, R is 255, etc. We have calculated MSE, PSNR for the whole image as well as for each plane in RGB
image. The results are shown below:
It is clearly visible from table 3. above that MSE and PSNR value of stego images are really good for images
as a whole as well as for individual planes. It may be noticed that MSE and PSNR values for blue plane are
little bit disappointing as compared to red and green plane values. This may be tolerated as human eyes are
less sensitive to blue color as compared to red and green colors.
9. 447
TABLE III: MSE AND PSNR VALUES FOR IMAGES WITH THEIR CORRESPONDING MSE AND SNR VALUES FOR EACH PLANE OF IMAGE
Sr.
No Image MSE PSNR
MSE-
Rplane
PSNR-
Rplane
MSE-
Gplane
PSNR-
Gplane
MSE-
Bplane
PSNR-
Bplane
1 Roses 1.266 47.108 0.204 55.030 0.249 54.161 0.819 49.036
2 Tiger 1.621 46.034 0.084 58.885 0.240 54.336 1.297 47.001
3 Horses 1.572 46.165 0.065 59.983 0.177 55.640 1.330 46.893
4 Red_Flowers 1.123 47.629 0.121 57.307 0.165 55.966 0.837 48.903
5 Policeman 1.507 46.350 0.055 60.722 0.065 59.984 1.386 46.712
6 Man_clrd_bkrnd 1.074 47.822 0.049 61.228 0.125 57.162 0.900 48.590
7 Eskimos 1.295 47.008 0.079 59.167 0.193 55.266 1.023 48.033
8 Swimmer 1.390 46.701 0.036 62.552 0.183 55.508 1.171 47.446
9 Water_lilies 1.472 46.451 0.111 57.668 0.346 52.736 1.015 48.067
10 Winter 1.765 45.663 0.200 55.126 0.492 51.215 1.074 47.821
11 Blue_hills 3.865 42.259 0.106 57.894 0.305 53.283 3.454 42.747
After inserting the secret message in each channel we have also improved the stego image by applying the 2k
correction to data carrier channel of the pixel (where k is no of bits replaced by secret data) as suggested by
Jae-Gil-Yu [40]. But here it was only limited to gray scale images. We have extended to the 2k correction for
RGB images and retrieved the unmatched results towards positive side in terms of PSNR value. The graphs
for proposed method with and without 2k correction are given below (Figure 7& 8) which itself reflects about
the strength of 2k correction.
Figure 7.Comparison of MSE values of Images with and without 2k correction
Figure 8.Comparison of PSNR values of Images with and without 2k correction
It can be noticed that 2k correction has really improved the values of MSE and PSNR almost for every image
and that too with good amount of difference. So, implementing 2k correction in RGB images seems to be
successful. Not only on MSE and PSNR parameter rather experiments have also been done on quality
parameter which shows promising results but due to lack of space the results for this parameters are not being
0.000
1.000
2.000
3.000
4.000
5.000
6.000
MSE(2k&W/o2k)Values
IMAGES
MSE(2k & W/o 2k)
Proosed 2K
Proosed W/o 2k
36.000
38.000
40.000
42.000
44.000
46.000
48.000
50.000
PSNR(2K&W/o2K)Values
Images
PSNR (2K & W/o 2K)
Proposed 2K
Proposed W/o 2k
10. 448
shown in this research paper. Further, secret data may be retrieved without using saliency map at the
receiving end. It is so because an indication has been embedded in the reflector channel which will guide the
user towards secret data bits in the pixels of the image. As the proposed scheme is using pseudo random
number generator using same seed (shared private key) at embedding as well as retrieving ends, so random
number generator will decide the same reflector channel in each pixel at both the ends. In this way retrieving
of secret data bits will be very much accurate at receiving end. In our experiments the accuracy in retrieving
secret bits at receiving end is almost 100%. Accuracy has been calculated by taking percentage of secret data
bits retrieved in same sequence in which it was embedded at transmitting end from total number of secret
data bits embedded. Experiments have been done on more than 40 test images from Berkeley image database.
VI. CONCLUSIONS
In this work we have designed an implemented an algorithm for computing the saliency for RGB images. It has
been shown that this saliency map is really better as compared to state of art methods based upon parameters of
precision, Recall etc. Further proposed algorithm is used for steganography in which saliency map is taken as
representative for Human visual system and based upon this model (saliency map) data is hidden in image
proportionately in salient and non-salient regions. Finally algorithm is checked for MSE and PSNR values which
really show promising results. Also 2k correction is applied to stego images to improve the visual results. These
MSE and P SNR values are also checked for individual planes of RGB image. The results are good for red and
green planes but somewhat discouraging for Blue plane which can be justified with well-known fact that only 2 %
cones of Human eye are sensitive to Blue color so this much tolerance can be accommodated.
ACKNOWLEDGEMENT
We wish to thank Punjab Technical University, Kapurthala to extend its all help for making this research
work possible. We have been provided infrastructure and other related facilities from Punjab Technical
University and its affiliated colleges. So contribution of Punjab Technical University to this research work is
really remarkable.
REFERENCES
[1] B. Wandell, Foundations of Vision, Sunderland, MA: Sinauer, 1995.
[2] L. Itti, C. Koch, and E. Niebur, “A model of saliency-based visual attention for rapid scene analysis,” in IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol. 20, pp. 1254-1259, Nov. 1998.
[3] L. Itti, and C. Koch, “Computational Modeling of visual attention,” in Nature Reviews, Neuroscience, vol. 2, pp.
194-203,Mar, 2001.
[4] T. Liu, N. Zheng, W. Ding and Z. Yuan, "Video attention: Learning to detect a salient object sequence," in Proc.
ICPR, pp.1-4, 2008.
[5] L. Itti, “Automatic Foveation for Video Compression Using a Neurobiological Model of Visual Attention,” in IEEE
Trans on Image Processing, vol. 13, no. 10, pp. 1304-1318, 2004.
[6] L. Itti, "Automatic attention-based prioritization of unconstrained video for compression," In Proc. SPIE Human
Vision and Electronic Imaging, vol. 5292, pp. 272-283, 2004.
[7] W. Lai, X. Gu, R. Wang, W. Ma, and H. Zhang, “A content-based bit allocation model for video streaming,” in proc.
IEEE international Conference on Multimedia and Expo (ICME), 2004.
[8] Z. Chen, G. Qiu, Y. Lu, L. Zhu, Q. Chen, X. Gu, and W. Charles, “Improving video coding at scene cuts using
attention based adaptive bit allocation,”, in Proc. ISCAS, pp. 3634-3638, 2007.
[9] .-F. Ma and H.-J. Zhang, "Contrast-based Image Attention Analysis by Using Fuzzy Growing", Proceedings of the
11th ACM International Conference on Multimedia, pp. 374-381, 2003.
[10] R. Achanta, F. Estrada, P. Wils, and S. Susstrunk, "Salient Region Detection and Segmentation", LNCS, Computer
Vision Systems,Springer BerlinlHeidelberg, pp. 66-75, 2008.
[11] X. Hou and L. Zhang, "Saliency Detection: A Spectral Residual Approach", IEEE Conference on Computer Vision
and Recognition 2007, pp. 1-8,2007.
[12] W. H. Ngau, L.-M. Ang, K. P. Seng, "Salient Region Detection Using Contrast-Based Saliency and Watershed
Segmentation", Proceedings of the 2nd International Conference on Computing and Informatics 2009 (lCOCI09),
pp. 475-479, 2009.
[13] U. Rajashekar, L. Cormack, and A. Boyik, "Visual Search: Structure from Noise", Proceedings of Eye Tracking
Research Applications Symposium", New Orleans, Los Angeles, USA, pp. 119-123,2002.
[14] W.-J. Won, S. Jeong, and M. Lee, "Road Traffic Sign Saliency Map Model", Proceedings of Image and Vision
Computing New Zealand 2007, New Zealand, pp. 91-96, 2007.