This paper presents a review of various OCR techniq
ues used in the automatic mail sorting process. A
complete description on various existing methods fo
r address block extraction and digit recognition th
at
were used in the literature is discussed. The objec
tive of this study is to provide a complete overvie
w about
the methods and techniques used by many researchers
for automating the mail sorting process in postal
service in various countries. The significance of Z
ip code or Pincode recognition is discussed.
Multi Resolution features of Content Based Image RetrievalIDES Editor
Many content based retrieval systems have been
proposed to manage and retrieve images on the basis of their
content. In this paper we proposed Color Histogram, Discrete
Wavelet Transform and Complex Wavelet Transform
techniques for efficient image retrieval from huge database.
Color Histogram technique is based on exact matching of
histogram of query image and database. Discrete Wavelet
transform technique retrieves images based on computation
of wavelet coefficients of subbands. Complex Wavelet
Transform technique includes computation of real and
imaginary part to extract the details from texture. The
proposed method is tested on COREL1000 database and
retrieval results have demonstrated a significant improvement
in precision and recall.
Comparison of various Image Registration Techniques with the Proposed Hybrid ...idescitation
Image Registration is termed as the method to
transform different forms of image data into one coordinate
system. Registration is a important part in image processing
which is used for matching the pictures which are obtained at
different time intervals or from various sensors. A broad range
of registration techniques have been developed for the various
types of image data. These techniques are independently
studied for many applications resulting in the large body of
result. Vision is the most advanced of human sensors, so
naturally images play one of the most important roles in
human perception. Image registration is one of the branches
encompassed by the diverse field of digital image processing.
Due to its importance in many application areas as well as
since its nature is complicated; image registration is now the
topic of much recent research. Registration algorithms tend
to compute transformations to set correspondence betweenthe two images. In this paper the survey is done on various
image registration techniques. Also the different techniques
are compared with the proposed system of the projec
Research Inventy : International Journal of Engineering and Scienceresearchinventy
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.
Wavelet-Based Color Histogram on Content-Based Image RetrievalTELKOMNIKA JOURNAL
The growth of image databases in many domains, including fashion, biometric, graphic design,
architecture, etc. has increased rapidly. Content Based Image Retrieval System (CBIR) is a technique used
for finding relevant images from those huge and unannotated image databases based on low-level features
of the query images. In this study, an attempt to employ 2nd level Wavelet Based Color Histogram (WBCH)
on a CBIR system is proposed. Image database used in this study are taken from Wang’s image database
containing 1000 color images. The experiment results show that 2nd level WBCH gives better precision
(0.777) than the other methods, including 1st level WBCH, Color Histogram, Color Co-occurrence Matrix,
and Wavelet texture feature. It can be concluded that the 2nd Level of WBCH can be applied to CBIR system.
Automated Colorization of Grayscale Images Using Texture DescriptorsIDES Editor
A novel example-based process for automated
colorization of grayscale images using texture descriptors
(ACTD) without any human intervention is proposed. By
analyzing a set of sample color images, coherent regions of
homogeneous textures are extracted. A multi-channel filtering
technique is used for texture-based image segmentation. For
each area of interest, state of the art texture descriptors are
then computed and stored, along with corresponding color
information. These texture descriptors and the color
information are used for colorization of a grayscale image with
similar textures. Given a grayscale image to be colorized, the
segmentation and feature extraction processes are repeated.
The texture descriptors are used to perform Content-Based
Image Retrieval (CBIR). The colorization process is performed
by chroma replacement. This research finds numerous
applications, ranging from classic film restoration and
enhancement, to adding valuable information into medical and
satellite imaging, and to enhance the detection of objects from
x-ray images at the airports.
Multi Resolution features of Content Based Image RetrievalIDES Editor
Many content based retrieval systems have been
proposed to manage and retrieve images on the basis of their
content. In this paper we proposed Color Histogram, Discrete
Wavelet Transform and Complex Wavelet Transform
techniques for efficient image retrieval from huge database.
Color Histogram technique is based on exact matching of
histogram of query image and database. Discrete Wavelet
transform technique retrieves images based on computation
of wavelet coefficients of subbands. Complex Wavelet
Transform technique includes computation of real and
imaginary part to extract the details from texture. The
proposed method is tested on COREL1000 database and
retrieval results have demonstrated a significant improvement
in precision and recall.
Comparison of various Image Registration Techniques with the Proposed Hybrid ...idescitation
Image Registration is termed as the method to
transform different forms of image data into one coordinate
system. Registration is a important part in image processing
which is used for matching the pictures which are obtained at
different time intervals or from various sensors. A broad range
of registration techniques have been developed for the various
types of image data. These techniques are independently
studied for many applications resulting in the large body of
result. Vision is the most advanced of human sensors, so
naturally images play one of the most important roles in
human perception. Image registration is one of the branches
encompassed by the diverse field of digital image processing.
Due to its importance in many application areas as well as
since its nature is complicated; image registration is now the
topic of much recent research. Registration algorithms tend
to compute transformations to set correspondence betweenthe two images. In this paper the survey is done on various
image registration techniques. Also the different techniques
are compared with the proposed system of the projec
Research Inventy : International Journal of Engineering and Scienceresearchinventy
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.
Wavelet-Based Color Histogram on Content-Based Image RetrievalTELKOMNIKA JOURNAL
The growth of image databases in many domains, including fashion, biometric, graphic design,
architecture, etc. has increased rapidly. Content Based Image Retrieval System (CBIR) is a technique used
for finding relevant images from those huge and unannotated image databases based on low-level features
of the query images. In this study, an attempt to employ 2nd level Wavelet Based Color Histogram (WBCH)
on a CBIR system is proposed. Image database used in this study are taken from Wang’s image database
containing 1000 color images. The experiment results show that 2nd level WBCH gives better precision
(0.777) than the other methods, including 1st level WBCH, Color Histogram, Color Co-occurrence Matrix,
and Wavelet texture feature. It can be concluded that the 2nd Level of WBCH can be applied to CBIR system.
Automated Colorization of Grayscale Images Using Texture DescriptorsIDES Editor
A novel example-based process for automated
colorization of grayscale images using texture descriptors
(ACTD) without any human intervention is proposed. By
analyzing a set of sample color images, coherent regions of
homogeneous textures are extracted. A multi-channel filtering
technique is used for texture-based image segmentation. For
each area of interest, state of the art texture descriptors are
then computed and stored, along with corresponding color
information. These texture descriptors and the color
information are used for colorization of a grayscale image with
similar textures. Given a grayscale image to be colorized, the
segmentation and feature extraction processes are repeated.
The texture descriptors are used to perform Content-Based
Image Retrieval (CBIR). The colorization process is performed
by chroma replacement. This research finds numerous
applications, ranging from classic film restoration and
enhancement, to adding valuable information into medical and
satellite imaging, and to enhance the detection of objects from
x-ray images at the airports.
APPLYING R-SPATIOGRAM IN OBJECT TRACKING FOR OCCLUSION HANDLINGsipij
Object tracking is one of the most important problems in computer vision. The aim of video tracking is to extract the trajectories of a target or object of interest, i.e. accurately locate a moving target in a video sequence and discriminate target from non-targets in the feature space of the sequence. So, feature descriptors can have significant effects on such discrimination. In this paper, we use the basic idea of many trackers which consists of three main components of the reference model, i.e., object modeling, object detection and localization, and model updating. However, there are major improvements in our system. Our forth component, occlusion handling, utilizes the r-spatiogram to detect the best target candidate. While spatiogram contains some moments upon the coordinates of the pixels, r-spatiogram computes region-based compactness on the distribution of the given feature in the image that captures richer features to represent the objects. The proposed research develops an efficient and robust way to keep tracking the object throughout video sequences in the presence of significant appearance variations and severe occlusions. The proposed method is evaluated on the Princeton RGBD tracking dataset considering sequences with different challenges and the obtained results demonstrate the effectiveness of the proposed method.
06 9237 it texture classification based edit putriIAESIJEECS
Automatic inspection systems become more importance for industries with high productive plans especially in texture industry. A novel approach to Local Binary Pattern (LBP) feature for texture classification is proposed in this system. At the first, the proposed Empirical Wavelet Transform (EWT) based texture classification is tested on gray scale and color images by using Brodatz texture images. The gray scale and color image is decomposed by EWT at 2 and 3 level of decomposition. LBP features are calculated for each empirical transformed image. Extracted features are given as input to the classification stage. K-NN classifier is used for classification stage. The result of the proposed system gives satisfactory classification accuracy of over 98% for all types of images.
A Framework for Curved Videotext Detection and ExtractionIJERA Editor
Proposed approach explores a new framework for curved video text detection and extraction. The algorithm first utilizes a Gaussian filter based Color Edge Enhancement followed by a Gray level Co-occurrence matrix feature extraction method for text detection. Secondly, a Connected Component filtering method is utilized to generate clear localization result and at last, a Round Scan method is performed to extract curved text and generate binary result for recognition by OCR. Experiments on various curved video data and Hua’s horizontal video text dataset shows the effectiveness and robustness of the proposed method.
Content Based Image Retrieval (CBIR) is one of the
most active in the current research field of multimedia retrieval.
It retrieves the images from the large databases based on images
feature like color, texture and shape. In this paper, Image
retrieval based on multi feature fusion is achieved by color and
texture features as well as the similarity measures are
investigated. The work of color feature extraction is obtained by
using Quadratic Distance and texture features by using Pyramid
Structure Wavelet Transforms and Gray level co-occurrence
matrix. We are comparing all these methods for best image
retrieval
PDE BASED FEATURES FOR TEXTURE ANALYSIS USING WAVELET TRANSFORMIJCI JOURNAL
In the present paper, a novel method of partial differential equation (PDE) based features for texture
analysis using wavelet transform is proposed. The aim of the proposed method is to investigate texture
descriptors that perform better with low computational cost. Wavelet transform is applied to obtain
directional information from the image. Anisotropic diffusion is used to find texture approximation from
directional information. Further, texture approximation is used to compute various statistical features.
LDA is employed to enhance the class separability. The k-NN classifier with tenfold experimentation is
used for classification. The proposed method is evaluated on Brodatz dataset. The experimental results
demonstrate the effectiveness of the method as compared to the other methods in the literature.
Colour Object Recognition using Biologically Inspired Modelijsrd.com
Human visual system can categorize objects rapidly and effortlessly despite the complexity and objective ambiguities of natural images. Despite the ease with which we see, visual categorization is an extremely difficult task for computers due to the variability of objects, such as scale, rotation, illumination, position and occlusion. Utilization of characteristics of biological systems for solving practical problems inevitably leads towards reducing the gap between manmade machines and live systems. Biologically motivated information processing has been an important area of scientific research for decades. This paper presents a Biologically Inspired Model which gives a promising solution to object categorization in colour space. The features are extracted in YCbCr colour space and classified by using SVM classifier. The framework has been applied to the image dataset taken from the Amsterdam Library of Object Images (ALOI). The proposed framework can successfully detect and classify the object categories with a good accuracy rate of about 91.3% for the Cb plane.
Parallax Effect Free Mosaicing of Underwater Video Sequence Based on Texture ...sipij
In this paper, we present feature-based technique for construction of mosaic image from underwater video
sequence, which suffers from parallax distortion due to propagation properties of light in the underwater
environment. The most of the available mosaic tools and underwater image mosaicing techniques yields
final result with some artifacts such as blurring, ghosting and seam due to presence of parallax in the input
images. The removal of parallax from input images may not reduce its effects instead it must be corrected
in successive steps of mosaicing. Thus, our approach minimizes the parallax effects by adopting an efficient
local alignment technique after global registration. We extract texture features using Centre Symmetric
Local Binary Pattern (CS-LBP) descriptor in order to find feature correspondences, which are used further
for estimation of homography through RANSAC. In order to increase the accuracy of global registration,
we perform preprocessing such as colour alignment between two selected frames based on colour
distribution adjustment. Because of existence of 100% overlap in consecutive frames of underwater video,
we select frames with minimum overlap based on mutual offset in order to reduce the computation cost
during mosaicing. Our approach minimizes the parallax effects considerably in final mosaic constructed
using our own underwater video sequences.
Beamforming with per antenna power constraint and transmit antenna selection ...sipij
In this paper, transmit beamforming and antenna selection techniques are presented for the Cooperative
Distributed Antenna System. Beamforming technique with minimum total weighted transmit power
satisfying threshold SINR and Per-Antenna Power constraints is formulated as a convex optimization
problem for the efficient performance of Distributed Antenna System (DAS). Antenna Selection technique is
implemented in this paper to select the optimum Remote Antenna Units from all the available ones. This
achieves the best compromise between capacity and system complexity. Dual polarized and Triple
Polarized systems are considered. Simulation results prove that by integrating Beamforming with DAS
enhances its performance. Also by using convex optimization in Antenna Selection enhances the
performance of multi polarized systems.
Global threshold and region based active contour model for accurate image seg...sipij
In this contribution, we develop a novel global threshold-based active contour model. This model deploys a new
edge-stopping function to control the direction of the evolution and to stop the evolving contour at weak or
blurred edges. An implementation of the model requires the use of selective binary and Gaussian filtering
regularized level set (SBGFRLS) method. The method uses either a selective local or global segmentation
property. It penalizes the level set function to force it to become a binary function. This procedure is followed by
using a regularisation Gaussian. The Gaussian filters smooth the level set function and stabilises the evolution
process. One of the merits of our proposed model stems from the ability to initialise the contour anywhere inside
the image to extract object boundaries. The proposed method is found to perform well, notably when the
intensities inside and outside the object are homogenous. Our method is applied with satisfactory results on
various types of images, including synthetic, medical and Arabic-characters images.
APPLYING R-SPATIOGRAM IN OBJECT TRACKING FOR OCCLUSION HANDLINGsipij
Object tracking is one of the most important problems in computer vision. The aim of video tracking is to extract the trajectories of a target or object of interest, i.e. accurately locate a moving target in a video sequence and discriminate target from non-targets in the feature space of the sequence. So, feature descriptors can have significant effects on such discrimination. In this paper, we use the basic idea of many trackers which consists of three main components of the reference model, i.e., object modeling, object detection and localization, and model updating. However, there are major improvements in our system. Our forth component, occlusion handling, utilizes the r-spatiogram to detect the best target candidate. While spatiogram contains some moments upon the coordinates of the pixels, r-spatiogram computes region-based compactness on the distribution of the given feature in the image that captures richer features to represent the objects. The proposed research develops an efficient and robust way to keep tracking the object throughout video sequences in the presence of significant appearance variations and severe occlusions. The proposed method is evaluated on the Princeton RGBD tracking dataset considering sequences with different challenges and the obtained results demonstrate the effectiveness of the proposed method.
06 9237 it texture classification based edit putriIAESIJEECS
Automatic inspection systems become more importance for industries with high productive plans especially in texture industry. A novel approach to Local Binary Pattern (LBP) feature for texture classification is proposed in this system. At the first, the proposed Empirical Wavelet Transform (EWT) based texture classification is tested on gray scale and color images by using Brodatz texture images. The gray scale and color image is decomposed by EWT at 2 and 3 level of decomposition. LBP features are calculated for each empirical transformed image. Extracted features are given as input to the classification stage. K-NN classifier is used for classification stage. The result of the proposed system gives satisfactory classification accuracy of over 98% for all types of images.
A Framework for Curved Videotext Detection and ExtractionIJERA Editor
Proposed approach explores a new framework for curved video text detection and extraction. The algorithm first utilizes a Gaussian filter based Color Edge Enhancement followed by a Gray level Co-occurrence matrix feature extraction method for text detection. Secondly, a Connected Component filtering method is utilized to generate clear localization result and at last, a Round Scan method is performed to extract curved text and generate binary result for recognition by OCR. Experiments on various curved video data and Hua’s horizontal video text dataset shows the effectiveness and robustness of the proposed method.
Content Based Image Retrieval (CBIR) is one of the
most active in the current research field of multimedia retrieval.
It retrieves the images from the large databases based on images
feature like color, texture and shape. In this paper, Image
retrieval based on multi feature fusion is achieved by color and
texture features as well as the similarity measures are
investigated. The work of color feature extraction is obtained by
using Quadratic Distance and texture features by using Pyramid
Structure Wavelet Transforms and Gray level co-occurrence
matrix. We are comparing all these methods for best image
retrieval
PDE BASED FEATURES FOR TEXTURE ANALYSIS USING WAVELET TRANSFORMIJCI JOURNAL
In the present paper, a novel method of partial differential equation (PDE) based features for texture
analysis using wavelet transform is proposed. The aim of the proposed method is to investigate texture
descriptors that perform better with low computational cost. Wavelet transform is applied to obtain
directional information from the image. Anisotropic diffusion is used to find texture approximation from
directional information. Further, texture approximation is used to compute various statistical features.
LDA is employed to enhance the class separability. The k-NN classifier with tenfold experimentation is
used for classification. The proposed method is evaluated on Brodatz dataset. The experimental results
demonstrate the effectiveness of the method as compared to the other methods in the literature.
Colour Object Recognition using Biologically Inspired Modelijsrd.com
Human visual system can categorize objects rapidly and effortlessly despite the complexity and objective ambiguities of natural images. Despite the ease with which we see, visual categorization is an extremely difficult task for computers due to the variability of objects, such as scale, rotation, illumination, position and occlusion. Utilization of characteristics of biological systems for solving practical problems inevitably leads towards reducing the gap between manmade machines and live systems. Biologically motivated information processing has been an important area of scientific research for decades. This paper presents a Biologically Inspired Model which gives a promising solution to object categorization in colour space. The features are extracted in YCbCr colour space and classified by using SVM classifier. The framework has been applied to the image dataset taken from the Amsterdam Library of Object Images (ALOI). The proposed framework can successfully detect and classify the object categories with a good accuracy rate of about 91.3% for the Cb plane.
Parallax Effect Free Mosaicing of Underwater Video Sequence Based on Texture ...sipij
In this paper, we present feature-based technique for construction of mosaic image from underwater video
sequence, which suffers from parallax distortion due to propagation properties of light in the underwater
environment. The most of the available mosaic tools and underwater image mosaicing techniques yields
final result with some artifacts such as blurring, ghosting and seam due to presence of parallax in the input
images. The removal of parallax from input images may not reduce its effects instead it must be corrected
in successive steps of mosaicing. Thus, our approach minimizes the parallax effects by adopting an efficient
local alignment technique after global registration. We extract texture features using Centre Symmetric
Local Binary Pattern (CS-LBP) descriptor in order to find feature correspondences, which are used further
for estimation of homography through RANSAC. In order to increase the accuracy of global registration,
we perform preprocessing such as colour alignment between two selected frames based on colour
distribution adjustment. Because of existence of 100% overlap in consecutive frames of underwater video,
we select frames with minimum overlap based on mutual offset in order to reduce the computation cost
during mosaicing. Our approach minimizes the parallax effects considerably in final mosaic constructed
using our own underwater video sequences.
Beamforming with per antenna power constraint and transmit antenna selection ...sipij
In this paper, transmit beamforming and antenna selection techniques are presented for the Cooperative
Distributed Antenna System. Beamforming technique with minimum total weighted transmit power
satisfying threshold SINR and Per-Antenna Power constraints is formulated as a convex optimization
problem for the efficient performance of Distributed Antenna System (DAS). Antenna Selection technique is
implemented in this paper to select the optimum Remote Antenna Units from all the available ones. This
achieves the best compromise between capacity and system complexity. Dual polarized and Triple
Polarized systems are considered. Simulation results prove that by integrating Beamforming with DAS
enhances its performance. Also by using convex optimization in Antenna Selection enhances the
performance of multi polarized systems.
Global threshold and region based active contour model for accurate image seg...sipij
In this contribution, we develop a novel global threshold-based active contour model. This model deploys a new
edge-stopping function to control the direction of the evolution and to stop the evolving contour at weak or
blurred edges. An implementation of the model requires the use of selective binary and Gaussian filtering
regularized level set (SBGFRLS) method. The method uses either a selective local or global segmentation
property. It penalizes the level set function to force it to become a binary function. This procedure is followed by
using a regularisation Gaussian. The Gaussian filters smooth the level set function and stabilises the evolution
process. One of the merits of our proposed model stems from the ability to initialise the contour anywhere inside
the image to extract object boundaries. The proposed method is found to perform well, notably when the
intensities inside and outside the object are homogenous. Our method is applied with satisfactory results on
various types of images, including synthetic, medical and Arabic-characters images.
Lossless image compression using new biorthogonal waveletssipij
Even though a large number of wavelets exist, one needs new wavelets for their specific applications. One
of the basic wavelet categories is orthogonal wavelets. But it was hard to find orthogonal and symmetric
wavelets. Symmetricity is required for perfect reconstruction. Hence, a need for orthogonal and symmetric
arises. The solution was in the form of biorthogonal wavelets which preserves perfect reconstruction
condition. Though a number of biorthogonal wavelets are proposed in the literature, in this paper four new
biorthogonal wavelets are proposed which gives better compression performance. The new wavelets are
compared with traditional wavelets by using the design metrics Peak Signal to Noise Ratio (PSNR) and
Compression Ratio (CR). Set Partitioning in Hierarchical Trees (SPIHT) coding algorithm was utilized to
incorporate compression of images.
A combined method of fractal and glcm features for mri and ct scan images cla...sipij
Fractal analysis has been shown to be useful in image processing for characterizing shape and gray-scale
complexity. The fractal feature is a compact descriptor used to give a numerical measure of the degree of
irregularity of the medical images. This descriptor property does not give ownership of the local image
structure. In this paper, we present a combination of this parameter based on Box Counting with GLCM
Features. This powerful combination has proved good results especially in classification of medical texture
from MRI and CT Scan images of trabecular bone. This method has the potential to improve clinical
diagnostics tests for osteoporosis pathologies.
An intensity based medical image registration using genetic algorithmsipij
Medical imaging plays a vital role to create images of human body for clinical purposes. Biomedical
imaging has taken a leap by entering into the field of image registration. Image registration integrates the
large amount of medical information embedded in the images taken at different time intervals and images
at different orientations. In this paper, an intensity-based real-coded genetic algorithm is used for
registering two MRI images. To demonstrate the efficiency of the algorithm developed, the alignment of the
image is altered and algorithm is tested for better performance. Also the work involves the comparison of
two similarity metrics, and based on the outcome the best metric suited for genetic algorithm is studied.
Speaker Identification From Youtube Obtained Datasipij
An efficient, and intuitive algorithm is presented for the identification of speakers from a long dataset (like
YouTube long discussion, Cocktail party recorded audio or video).The goal of automatic speaker
identification is to identify the number of different speakers and prepare a model for that speaker by
extraction, characterization and speaker-specific information contained in the speech signal. It has many
diverse application specially in the field of Surveillance , Immigrations at Airport , cyber security ,
transcription in multi-source of similar sound source, where it is difficult to assign transcription arbitrary.
The most commonly speech parameterization used in speaker verification, K-mean, cepstral analysis, is
detailed. Gaussian mixture modeling, which is the speaker modeling technique is then explained. Gaussian
mixture models (GMM), perhaps the most robust machine learning algorithm has been introduced to
examine and judge carefully speaker identification in text independent. The application or employment of
Gaussian mixture models for monitoring & Analysing speaker identity is encouraged by the familiarity,
awareness, or understanding gained through experience that Gaussian spectrum depict the characteristics
of speaker's spectral conformational pattern and remarkable ability of GMM to construct capricious
densities after that we illustrate 'Expectation maximization' an iterative algorithm which takes some
arbitrary value in initial estimation and carry on the iterative process until the convergence of value is
observed We have tried to obtained 85 ~ 95% of accuracy using speaker modeling of vector quantization
and Gaussian Mixture model ,so by doing various number of experiments we are able to obtain 79 ~ 82%
of identification rate using Vector quantization and 85 ~ 92.6% of identification rate using GMM modeling
by Expectation maximization parameter estimation depending on variation of parameter.
Offline handwritten signature identification using adaptive window positionin...sipij
The paper presents to address this challenge, we have proposed the use of Adaptive Window Positioning
technique which focuses on not just the meaning of the handwritten signature but also on the individuality
of the writer. This innovative technique divides the handwritten signature into 13 small windows of size nxn
(13x13). This size should be large enough to contain ample information about the style of the author and
small enough to ensure a good identification performance. The process was tested with a GPDS dataset
containing 4870 signature samples from 90 different writers by comparing the robust features of the test
signature with that of the user’s signature using an appropriate classifier. Experimental results reveal that
adaptive window positioning technique proved to be the efficient and reliable method for accurate
signature feature extraction for the identification of offline handwritten signatures .The contribution of this
technique can be used to detect signatures signed under emotional duress
Feature selection approach in animal classificationsipij
In this paper, we propose a model for automatic classification of Animals using different classifiers Nearest
Neighbour, Probabilistic Neural Network and Symbolic. Animal images are segmented using maximal
region merging segmentation. The Gabor features are extracted from segmented animal images.
Discriminative texture features are then selected using the different feature selection algorithm like
Sequential Forward Selection, Sequential Floating Forward Selection, Sequential Backward Selection and
Sequential Floating Backward Selection. To corroborate the efficacy of the proposed method, an
experiment was conducted on our own data set of 25 classes of animals, containing 2500 samples. The
data set has different animal species with similar appearance (small inter-class variations) across different
classes and varying appearance (large intra-class variations) within a class. In addition, the images of
flowers are of different poses, with cluttered background under different lighting and climatic conditions.
Experiment results reveal that Symbolic classifier outperforms Nearest Neighbour and Probabilistic Neural
Network classifiers.
A Novel Uncertainty Parameter SR ( Signal to Residual Spectrum Ratio ) Evalua...sipij
Usually, hearing impaired people use hearing aids which are implemented with speech enhancement
algorithms. Estimation of speech and estimation of nose are the components in single channel speech
enhancement system. The main objective of any speech enhancement algorithm is estimation of noise power
spectrum for non stationary environment. VAD (Voice Activity Detector) is used to identify speech pauses
and during these pauses only estimation of noise. MMSE (Minimum Mean Square Error) speech
enhancement algorithm did not enhance the intelligibility, quality and listener fatigues are the perceptual
aspects of speech. Novel evaluation approach SR (Signal to Residual spectrum ratio) based on uncertainty
parameter introduced for the benefits of hearing impaired people in non stationary environments to control
distortions. By estimation and updating of noise based on division of original pure signal into three parts
such as pure speech, quasi speech and non speech frames based on multiple threshold conditions. Different
values of SR and LLR demonstrate the amount of attenuation and amplification distortions. The proposed
method will compared with any one method WAT(Weighted Average Technique) Hence by using
parameters SR (signal to residual spectrum ratio) and LLR (log like hood ratio), MMSE (Minim Mean
Square Error) in terms of segmented SNR and LLR.
Robust content based watermarking algorithm using singular value decompositio...sipij
Nowadays, image content is frequently subject to different malicious manipulations. To protect images
from this illegal manipulations computer science community have recourse to watermarking techniques. To
protect digital multimedia content we need just to embed an invisible watermark into images which
facilitate the detection of different manipulations, duplication, illegitimate distributions of these images. In
this work a robust watermarking technique is presented that embedding invisible watermarks into colour
images the singular value decomposition bloc by bloc of a robust transform of images that is the Radial
symmetry transform. Each bit of the watermark is inserted in a bloc of eight pixels large of the blue
channel a high singular value of the corresponding bloc into the radial symmetry map. We justified the
insertion in the blue channel by our feeble sensibility to perturbations in this colour channel of images. We
present also results obtained with different tests. We had tested the imperceptibility of the mark using this
approach and also its robustness face to several attacks.
Application of parallel algorithm approach for performance optimization of oi...sipij
This paper gives a detailed study on the performance of image filter algorithm with various parameters
applied on an image of RGB model. There are various popular image filters, which consumes large amount
of computing resources for processing. Oil paint image filter is one of the very interesting filters, which is
very performance hungry. Current research tries to find improvement in oil paint image filter algorithm by
using parallel pattern library. With increasing kernel-size, the processing time of oil paint image filter
algorithm increases exponentially. I have also observed in various blogs and forums, the questions for
faster oil paint have been asked repeatedly.
Contrast enhancement using various statistical operations and neighborhood pr...sipij
Histogram Equalization is a simple and effective contrast enhancement technique. In spite of its popularity
Histogram Equalization still have some limitations –produces artifacts, unnatural images and the local
details are not considered, therefore due to these limitations many other Equalization techniques have been
derived from it with some up gradation. In this proposed method statistics play an important role in image
processing, where statistical operations is applied to the image to get the desired result such as
manipulation of brightness and contrast. Thus, a novel algorithm using statistical operations and
neighborhood processing has been proposed in this paper where the algorithm has proven to be effective in
contrast enhancement based on the theory and experiment.
A voting based approach to detect recursive order number of photocopy documen...sipij
Photocopy documents are very common in our normal life. People are permitted to carry and present
photocopied documents to avoid damages to the original documents. But this provision is misused for
temporary benefits by fabricating fake photocopied documents. Fabrication of fake photocopied document
is possible only in 2nd and higher order recursive order of photocopies. Whenever a photocopied document
is submitted, it may be required to check its originality. When the document is 1st order photocopy, chances
of fabrication may be ignored. On the other hand when the photocopy order is 2nd or above, probability of
fabrication may be suspected. Hence when a photocopy document is presented, the recursive order number
of photocopy is to be estimated to ascertain the originality. This requirement demands to investigate
methods to estimate order number of photocopy. In this work, a voting based approach is used to detect the
recursive order number of the photocopy document using probability distributions exponential, extreme
values and lognormal distributions is proposed. A detailed experimentation is performed on a generated
data set and the method exhibits efficiency close to 89%.
Image retrieval and re ranking techniques - a surveysipij
There is a huge amount of research work focusing on the searching, retrieval and re-ranking of images in
the image database. The diverse and scattered work in this domain needs to be collected and organized for
easy and quick reference.
Relating to the above context, this paper gives a brief overview of various image retrieval and re-ranking
techniques. Starting with the introduction to existing system the paper proceeds through the core
architecture of image harvesting and retrieval system to the different Re-ranking techniques. These
techniques are discussed in terms of approaches, methodologies and findings and are listed in tabular form
for quick review.
IDENTIFICATION OF SUITED QUALITY METRICS FOR NATURAL AND MEDICAL IMAGESsipij
To assess quality of the denoised image is one of the important task in image denoising application.
Numerous quality metrics are proposed by researchers with their particular characteristics till today. In
practice, image acquisition system is different for natural and medical images. Hence noise introduced in
these images is also different in nature. Considering this fact, authors in this paper tried to identify the
suited quality metrics for Gaussian, speckle and Poisson corrupted natural, ultrasound and X-ray images
respectively. In this paper, sixteen different quality metrics from full reference category are evaluated with
respect to noise variance and suited quality metric for particular type of noise is identified. Strong need to
develop noise dependent quality metric is also identified in this work.
International Journal of Engineering and Science Invention (IJESI)inventionjournals
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.
Introduction to image processing and pattern recognitionSaibee Alam
this power point presentation provides a brief introduction to image processing and pattern recognition and its related research papers including conclusion
A Novel Framework For Numerical Character Recognition With Zoning Distance Fe...IJERD Editor
Advancements of Computer technology has made every organization to implement the automatic processing systems for its activities. One of the examples is the recognition of handwritten characters, which has always been a challenging task in image processing and pattern recognition. In this paper we propose Zone based features for recognition of the handwritten characters. In this zoning approach a digit image is divided into 8x8 zones and centre pixel is computed for each zone. This procedure is sequentially repeated for entire zone. Finally features are extracted for classification and recognition.
A NOVEL DATA DICTIONARY LEARNING FOR LEAF RECOGNITIONsipij
Automatic leaf recognition via image processing has been greatly important for a number of professionals, such as botanical taxonomic, environmental protectors, and foresters. Learn an over-complete leaf dictionary is an essential step for leaf image recognition. Big leaf images dimensions and training images number is facing of fast and complete data leaves dictionary. In this work an efficient approach applies to construct over-complete data leaves dictionary to set of big images diminutions based on sparse representation. In the proposed method a new cropped-contour method has used to crop the training image. The experiments are testing using correlation between the sparse representation and data dictionary and with focus on the computing time.
Empirical Coding for Curvature Based Linear Representation in Image Retrieval...iosrjce
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.
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.
Multilinear Kernel Mapping for Feature Dimension Reduction in Content Based M...ijma
In the process of content-based multimedia retrieval, multimedia information is processed in order to
obtain descriptive features. Descriptive representation of features, results in a huge feature count, which
results in processing overhead. To reduce this descriptive feature overhead, various approaches have been
used to dimensional reduction, among which PCA and LDA are the most used methods. However, these
methods do not reflect the significance of feature content in terms of inter-relation among all dataset
features. To achieve a dimension reduction based on histogram transformation, features with low
significance can be eliminated. In this paper, we propose a feature dimensional reduction approaches to
achieve the dimension reduction approach based on a multi-linear kernel (MLK) modeling. A benchmark
dataset for the experimental work is taken and the proposed work is observed to be improved in analysis in
comparison to the conventional system.
Human Re-identification with Global and Local Siamese Convolution Neural NetworkTELKOMNIKA JOURNAL
Human re-identification is an important task in surveillance system to determine whether the same human re-appears in multiple cameras with disjoint views. Mostly, appearance based approaches are used to perform human re-identification task because they are less constrained than biometric based approaches. Most of the research works apply hand-crafted feature extractors and then simple matching methods are used. However, designing a robust and stable feature requires expert knowledge and takes time to tune the features. In this paper, we propose a global and local structure of Siamese Convolution Neural Network which automatically extracts features from input images to perform human re-identification task. Besides, most of the current human re-identification tasks in single-shot approaches do not consider occlusion issue due to lack of tracking information. Therefore, we apply a decision fusion technique to combine global and local features for occlusion cases in single-shot approaches.
Application of Image Retrieval Techniques to Understand Evolving Weatherijsrd.com
Multispectral satellite images provide valuable information to understand the evolution of various weather systems such as tropical cyclones, shifting of intra tropical convergence zone, moments of various troughs etc., accurate prediction and estimation will save live and property. This work will deal with the development of an application which will enable users to search an image from database using either gray level, texture and shape features for meteorological satellite image retrieval .Gray level feature is extracted using histogram method. The Texture feature is extracted using gray level co-occurrence method and wavelet approach. The shape feature vector is extracted using morphological operations. The similarity between query image and database images is calculated using Euclidian distance. The performance of the system is evaluated using precision
Facial image retrieval on semantic features using adaptive mean genetic algor...TELKOMNIKA JOURNAL
The emergence of larger databases has made image retrieval techniques an essential component and has led to the development of more efficient image retrieval systems. Retrieval can either be content or text-based. In this paper, the focus is on the content-based image retrieval from the FGNET database. Input query images are subjected to several processing techniques in the database before computing the squared Euclidean distance (SED) between them. The images with the shortest Euclidean distance are considered as a match and are retrieved. The processing techniques involve the application of the median modified Weiner filter (MMWF), extraction of the low-level features using histogram-oriented gradients (HOG), discrete wavelet transform (DWT), GIST, and Local tetra pattern (LTrP). Finally, the features are selected using Adaptive Mean Genetic Algorithm (AMGA). In this study, the average PSNR value obtained after applying the Wiener filter was 45.29. The performance of the AMGA was evaluated based on its precision, F-measure, and recall, and the obtained average values were respectively 0.75, 0.692, and 0.66. The performance matrix of the AMGA was compared to those of particle swarm optimization algorithm (PSO) and genetic algorithm (GA) and found to perform better; thus, proving its efficiency.
Similar to Review of ocr techniques used in automatic mail sorting of postal envelopes (20)
Epistemic Interaction - tuning interfaces to provide information for AI supportAlan Dix
Paper presented at SYNERGY workshop at AVI 2024, Genoa, Italy. 3rd June 2024
https://alandix.com/academic/papers/synergy2024-epistemic/
As machine learning integrates deeper into human-computer interactions, the concept of epistemic interaction emerges, aiming to refine these interactions to enhance system adaptability. This approach encourages minor, intentional adjustments in user behaviour to enrich the data available for system learning. This paper introduces epistemic interaction within the context of human-system communication, illustrating how deliberate interaction design can improve system understanding and adaptation. Through concrete examples, we demonstrate the potential of epistemic interaction to significantly advance human-computer interaction by leveraging intuitive human communication strategies to inform system design and functionality, offering a novel pathway for enriching user-system engagements.
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024Albert Hoitingh
In this session I delve into the encryption technology used in Microsoft 365 and Microsoft Purview. Including the concepts of Customer Key and Double Key Encryption.
UiPath Test Automation using UiPath Test Suite series, part 4DianaGray10
Welcome to UiPath Test Automation using UiPath Test Suite series part 4. In this session, we will cover Test Manager overview along with SAP heatmap.
The UiPath Test Manager overview with SAP heatmap webinar offers a concise yet comprehensive exploration of the role of a Test Manager within SAP environments, coupled with the utilization of heatmaps for effective testing strategies.
Participants will gain insights into the responsibilities, challenges, and best practices associated with test management in SAP projects. Additionally, the webinar delves into the significance of heatmaps as a visual aid for identifying testing priorities, areas of risk, and resource allocation within SAP landscapes. Through this session, attendees can expect to enhance their understanding of test management principles while learning practical approaches to optimize testing processes in SAP environments using heatmap visualization techniques
What will you get from this session?
1. Insights into SAP testing best practices
2. Heatmap utilization for testing
3. Optimization of testing processes
4. Demo
Topics covered:
Execution from the test manager
Orchestrator execution result
Defect reporting
SAP heatmap example with demo
Speaker:
Deepak Rai, Automation Practice Lead, Boundaryless Group and UiPath MVP
Securing your Kubernetes cluster_ a step-by-step guide to success !KatiaHIMEUR1
Today, after several years of existence, an extremely active community and an ultra-dynamic ecosystem, Kubernetes has established itself as the de facto standard in container orchestration. Thanks to a wide range of managed services, it has never been so easy to set up a ready-to-use Kubernetes cluster.
However, this ease of use means that the subject of security in Kubernetes is often left for later, or even neglected. This exposes companies to significant risks.
In this talk, I'll show you step-by-step how to secure your Kubernetes cluster for greater peace of mind and reliability.
DevOps and Testing slides at DASA ConnectKari Kakkonen
My and Rik Marselis slides at 30.5.2024 DASA Connect conference. We discuss about what is testing, then what is agile testing and finally what is Testing in DevOps. Finally we had lovely workshop with the participants trying to find out different ways to think about quality and testing in different parts of the DevOps infinity loop.
Transcript: Selling digital books in 2024: Insights from industry leaders - T...BookNet Canada
The publishing industry has been selling digital audiobooks and ebooks for over a decade and has found its groove. What’s changed? What has stayed the same? Where do we go from here? Join a group of leading sales peers from across the industry for a conversation about the lessons learned since the popularization of digital books, best practices, digital book supply chain management, and more.
Link to video recording: https://bnctechforum.ca/sessions/selling-digital-books-in-2024-insights-from-industry-leaders/
Presented by BookNet Canada on May 28, 2024, with support from the Department of Canadian Heritage.
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...UiPathCommunity
💥 Speed, accuracy, and scaling – discover the superpowers of GenAI in action with UiPath Document Understanding and Communications Mining™:
See how to accelerate model training and optimize model performance with active learning
Learn about the latest enhancements to out-of-the-box document processing – with little to no training required
Get an exclusive demo of the new family of UiPath LLMs – GenAI models specialized for processing different types of documents and messages
This is a hands-on session specifically designed for automation developers and AI enthusiasts seeking to enhance their knowledge in leveraging the latest intelligent document processing capabilities offered by UiPath.
Speakers:
👨🏫 Andras Palfi, Senior Product Manager, UiPath
👩🏫 Lenka Dulovicova, Product Program Manager, UiPath
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...Ramesh Iyer
In today's fast-changing business world, Companies that adapt and embrace new ideas often need help to keep up with the competition. However, fostering a culture of innovation takes much work. It takes vision, leadership and willingness to take risks in the right proportion. Sachin Dev Duggal, co-founder of Builder.ai, has perfected the art of this balance, creating a company culture where creativity and growth are nurtured at each stage.
Monitoring Java Application Security with JDK Tools and JFR Events
Review of ocr techniques used in automatic mail sorting of postal envelopes
1. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
REVIEW OF OCR TECHNIQUES USED IN
AUTOMATIC MAIL SORTING OF POSTAL ENVELOPES
Dr.R.Radha and R.R.Aparna
Research Department of Computer science, SDNB Vaishnav College,
Affiliated to Madras University, Chennai, India.
ABSTRACT
This paper presents a review of various OCR techniques used in the automatic mail sorting process. A
complete description on various existing methods for address block extraction and digit recognition that
were used in the literature is discussed. The objective of this study is to provide a complete overview about
the methods and techniques used by many researchers for automating the mail sorting process in postal
service in various countries. The significance of Zip code or Pincode recognition is discussed.
KEYWORDS
Neural Network, OCR, Pincode, Zipcode, Segmentation, Recognition, Address Block, Feature Extraction,
Back propagation.
1. INTRODUCTION
The mechanization of mail sorting started in the year 1920. The first sorting machine was put into
operation in the year 1950. From 1982 onwards developments were made in Optical Character
Reader (OCR) which reads the mail piece destination address and Pincode and prints a Barcode
for further sorting. This study would guide the researchers with existing methods in developing
new OCR software to automate the mail sorting process in developing countries at a lesser cost.
To automate the sorting process initially the destination address block (DAB) has to be
segmented from the envelope image, then the Zip code has to be located. The extracted Zip code
is segmented into individual digits and recognition is performed. After recognition a relevant bar
code pertaining the destination address is printed (Figure 1). This paper discusses the various
methods used by many researches in the literature for performing DAB segmentation and
character recognition. Section 2 gives the descriptions of various methods used for DAB
segmentation. Section 3 discusses the various techniques used for Zip code recognition. Section 4
gives the performance comparison of different approaches proposed by different researchers in
the literature. In Section 5 conclusion and future work is discussed.
Figure 1: Procedure for Automatic mail sorting
DOI : 10.5121/sipij.2013.4504
45
2. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
2. REVIEW OF VARIOUS METHODS USED FOR DAB SEGMENTATION
Menoti et al. [23] have developed a DAB Segmentation algorithm based on feature selection in
wavelet space. Experiments were performed on Brazilian postal envelopes. The images were
decomposed into wavelet space using Mallat decomposition with Haar basis to identify the salient
points. The four features based on low (L) and high (H) frequency (LL, LH, HL, HH) were
obtained. 440 images with different layout and backgrounds were used. 85% accuracy was
achieved.
Wolf and Platt [40] have used Convolutional locator networks (CLN) to perform address block
location on machine printed mail pieces that belongs to the US postal service(USPS). Varied
shapes,sizes and justifications of DAB were considered. Instead of obtaining low level features
high level abstract features of an address block (AB) was obtained using CLN.CLN detects the
corners of AB.Explicit rules or models for extracting DAB was not used. The obtained features
were converted to object hypothesis. The output of CLN gives a four feature map each
representing a corner. Based on the intensity of the pixel in the feature map the corners were
located. CLN was constructed with 3 layers, using back propagation algorithm . Feature maps of
CLN were converted into address block location (ABL) candidates. 500 images were tested.
98.2% accuracy was obtained.
Legal-Ayala and Facon [17] have used a learning based approach for postal envelope address
block segmentation. Many features like pixel gray level values, mean, variance, skew and kurtosis
were considered. Learning followed by segmentation was performed. 200 samples of Brazilian
postal envelopes were taken. 10 samples were randomly chosen and submitted for learning stage.
The 10 sample images and their ideal images (expected output images) were compared. The
learning process computes all features of every pixel in both the images. The features were
extracted from each pair of pixels that belongs to both the images along their neighbourhood.
These extracted features were stored in the classification array. The other images were tested and
DAB was segmented based on the stored classification array by using the K-Nearest Neighbour
(KNN) and Euclidean distance measure. 98.6% accuracy was achieved.
Gaceb et al. [8] have implemented a robust approach of address block localization in business
mail by graph colouring technique. Graph colouring was performed to automatically separate the
elements into homogeneous groups. Training was performed using b-colouring technique. AB
was located using hierarchical graph colouring (HGC) and pyramidal data organization. The
colouring was performed by applying different colour to the adjacent nodes in the graph. If a
vertex is surrounded by all the specified number of colours then that node was referred as
dominant node. Training was performed based on this b-colouring. This technique separates the
dominant node that ensures a great inter class disparity. HGC was followed as it largely makes
use of all the levels of pyramidal structure to group the objects of same nature. 98% accuracy on
750 samples was obtained.
Eiterer et al. [5] have performed postal envelope address block location by fractal based
approach. Samples were taken from Brazilian post office. The fractal dimension of each pixel of
postal envelope was computed using 2D variation. The fractal dimensions technique measures the
roughness of a surface. Clustering was performed by K-means algorithm. After computing the
fractal dimension for each image pixel, the resulting fractal image was then clustered by K-means
into three clusters as background, post mark and stamp. The 2D variation procedure for these
three clusters of neighbour window size were chosen as (r=3,5, r=3,5,7, r=3,5,7,9). 200 postal
envelopes were taken. They obtained an accuracy of 97% for segmenting AB, 66% for stamp and
92% for post mark by considering a region box of size r=3,5,7,9.
46
3. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
Yu et al. [43] have developed an algorithm based on heuristics for identifying the address written
on light coloured background on an embossed label in the specified window of the envelope. The
bottom up approach was used to build a pyramid model. Modified Ostu’s method and connected
component analysis (CCA) were used and by applying some heuristics the DAB was located. 109
samples (53 IBM magazine envelopes and 56 other envelopes) were taken. 72% and 93%
accuracy was obtained.
Akira Yonekura and Facon [1] have used 2D histogram and morphological clustering based on
Watershed transform to segment the DAB. Digitized image and its filtered version were taken.
2D histogram contains some statistical information about the gray values of the pixel at the same
location in both the images. The morphological gray scale dual reconstruction preserves the peaks
and valleys. The envelope image was segmented into 3 classes background, stamps and address
blocks. 300 Brazil postal envelopes having complex background were taken as samples and the
accuracy for background, stamp and AB was 90%, 25% and 75% respectively.
Facon et al. [6] have proposed an approach based on lacunarity to locate the AB in Brazil postal
envelopes. Lacunarity depends on mean and variance for the window size throughout the images.
It is a multi-scale measure describing the distribution of gaps within the texture. Region growing
technique was applied to reconstruct the semantic objects like stamps, postmarks and address
blocks. The features were extracted using lacunarity and normalized using non-linear transform.
Threshold was applied to identify the objects. 200 samples were used and 93.5% accuracy was
achieved.
Xue et al. [41] have presented a new method to locate and interpret the DAB on handwritten
Chinese envelopes. The features were extracted using geometric features, positional information,
area and number of components. A new bottom up method was used to locate the DAB. It first
extracts the connected component (CC) and then classifies them into text blocks and non-text
blocks based on the obtained features. The segmented text blocks were merged to obtain the
address block. 80% accuracy was achieved.
Idrissi [14] has used ensemble clustering techniques for address block segmentation. Cluster
ensemble attempts to find more accurate and robust clustering results by combining the clustering
results of a single or multiple clustering algorithms. The samples were segmented using single
clustering algorithm and some features were extracted. The ensemble algorithm was developed by
using the base clusters or many single clustering algorithms and then the clusters were combined.
The graphics were removed from the image by using heuristics and by setting threshold. Single
clustering algorithms like expected maximization (EM), single linkage, complete linkage were
used. The cluster with the highest confidence score was segmented. The software was developed
for the user to choose the appropriate clustering algorithm or combined algorithm. The accuracy
of the algorithm can be tested by supplying different features for same algorithm or random data.
2000 samples form USPS and 1600 parcel images from Dutch sorting centre was used. Best
results were obtained. It was proved that ensemble technique improved the accuracy of clustering
process but resulted in increasing the computational time.
Govindaraju and Tulyakov [10] have performed DAB location by contour clustering based on
heuristics. The CC was clustered based on the extracted features and contours. They achieved
success by using agglomerative type clustering algorithms. The algorithm was similar to
minimum spanning tree. Threshold setting was used and ranking was performed to classify the
clusters on DAB. 1684 images from CEDAR data base were used. 1/3rd of it was tested and the
DAB was correctly located on 272 images. They noted 50% of accuracy and an increased
performance.
47
4. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
Wang [39] has located the DAB on images of complex mail pieces, based on the size and the
pixel density feature (PDF) of bounding box (BB) and heuristics. If the size is less than 10 x 10
pixels or greater than 100 x 100 pixels then those objects were removed considering them to be
noise or graphics. By using positional information and region growing the DAB was segmented.
100 Chinese test image envelopes were used.90.7% accuracy was achieved.
Bochnia and Facon [3] have performed address block extraction of magazine envelopes by
applying multiple thresholds. Address was written on a label on the envelope. Each image was
applied with threshold values that were modified from Pun, Johnnsen, Bille and Li algorithms.
The final segmentation was obtained from the highest threshold. 70 images of News Paper
envelopes having complex background were used and 80% accuracy was achieved.
3. REVIEW OF METHODS USED FOR DIGIT RECOGNITION
Gao and Jin [9] have developed a vision based fast postal envelope identification system for
isolated machine printed Chinese postal envelopes. Since the envelopes were of predefined sizes,
the DAB was located by setting a threshold value. Address Line were segmented into sub
segments using projection profile (PP). The segmentation path network was constructed in these
sub segments and the best segmentation point was scored by the classifier. Feature extraction was
performed by using 2D Gabor feature and gradient feature on binary and gray scale images. 256
dimensions Gabor feature and 516 dimension gradient features were used for recognition.
Classification was performed using Euclidean distance. Linear discriminative analysis (LDA) was
used for reducing the feature dimension. The Gabor bin obtained 98.92% for post code with a
font size no lesser than 10.5. The experiments were performed on 761 mail images which contain
25060 characters. Average of 32.9 characters in 81 milliseconds has achieved a recognition
accuracy of 98.7%. Accuracy was highly dependent on font size.
Pal et al. [25] have experimented with Indian postal letters to segment the DAB and recognize the
Pin code. Vertical and horizontal smoothing was performed using Run length smoothing
algorithm (RLSA) to decompose the image into blocks. Based on pixel intensity, the text block
(DAB) and non-text block (stamp, seal) were identified. Using positional information the DAB
was segmented. The Pincode was localized by CCA assuming the number of digits in the pin
code and based on some length and width measurements. The extracted pin code digits were
normalized using aspect ratio adaptive normalization (ARAN). The numerals were recognized
using two stage Multi-Layer Perceptron (MLP) classifier. The handwritten address contains
Bangla and Arabic numerals. Back propagation (BP) algorithm was used. The input layer
contains 784 neurons, output layer contains 16 neurons and hidden layer contains 400 neurons.
The experiments were conducted on 2860 postal images, accuracy for DAB segmentation was
98.55%, Pincode box recognition was 97.64%. Recognition of 15096 numerals samples were
taken, 8690 were used for training (Bangla and Arabic) and 6406 was used for testing. Bangla
digit recognition accuracy was 94.13% and Arabic numeral recognition accuracy 93.00%.
Roy et al. [28] have proposed a recognition method for recognizing the 6 digit Indian handwritten
pin code in the DAB of postal documents that were written in English, Bangla and Hindi (Multiscript). The DAB was extracted using the method proposed by [25]. From the extracted DAB, the
pin code was detected using positional information and segmented using Water Reservoir
method. To the segmented numerals Modified quadratic discriminant function (MQDF) was
applied. They used two feature set as proposed by [15]. Recognition was performed using
dynamic programming. 16300 pin code samples were taken which contains 2692 Bangla, 8184
English, 5424 Devanagri. 94.14% of recognition accuracy was obtained.
48
5. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
Sinha [31] has introduced a scheme for locating and recognizing words based on over
segmentation followed by dynamic programming. The USPS postal samples containing cursive
script postal address were used. ABL was performed by differentiating the foreground and
background details by using Hidden Markov Model (HMM) based technique for ZIP location.
But it was performed with minimum confidence. Hence recognition was performed. Zip code
location and recognition were performed using shortest path problems. Three MLP classifiers
with BP algorithm were used. The general classifier (G) was constructed with 108 units in input
layer, 120 units in the hidden layer and the output layer comprises 43 units. The digit classifier
(D) contains 108 units in the input layer, 50 units in the Hidden layer and 11 units in the output
layer. The alphabet classifier (A) contains 108 units in the input layer, 80 units in the hidden layer
and 27 units in the output layer. 562 images were tested. 98% accuracy was obtained for Zip code
location and 95% accuracy for Zip code recognition was obtained.
Ibrahim [12] has applied domain knowledge to the recognition of Zip code. He has developed the
system by understanding the structure of Zip code by recognizing the first digit. It involves 2
stages, segmentation followed by recognition. He has constructed 26 MLP networks. Initially 9
MLP networks were constructed for digit (0-9) that identifies the state. Then based on the
metaclass (second digit and other combinations), other 17 MLP networks were designed. 66214
digit samples were taken. 16000 was used for training and tested with 435 digits that belonged to
CEDAR CDROM-1.His model relies on the prior knowledge of destination state. 88.76%
accuracy was achieved.
Kimura et al. [15] have performed handwritten Zip code recognition using Lexicon free word
recognition algorithm. Address block samples were taken form USPS database. The Zip Codes
were extracted manually. Local Chain Code histogram for character contour was used as feature
vector. The BB of each digit in the Zip code was divided into 7 X 7 blocks. In each block the
Chain code histogram was calculated by using 4 directions. The size was down sampled to 4 X 4
blocks using Gaussian filters and 64 dimensions were obtained. The same way 400 dimensions
were obtained by 9 X9 blocks by using 16 directions.5590 samples were taken and they obtained
an accuracy of 99.2%.The same method was applied by [3] but for Indian postal envelope
containing English, Bangla and Hindi characters.
Le Cun et al. [16] have used Multilayer networks for performing handwritten Zip code
recognition. BP was used to recognize digits. USPS samples were used. The images were
normalized using linear transformations. 9298 segmented numerals were taken, 7291 handwritten
digits and 2549 printed digits were used for training.2007 handwritten digits and 7000 printed
digits were tested using MLP which contains 4 hidden layers. After 30 training phases accuracy
of 97% was obtained with 3% rejection rate.
Vajda et al. [36] have proposed a new method for automatic recognition of Bangla and English
numeral. Non-symmetric half plane (NSHP-HMM) was used for digit recognition. DAB was
extracted using the method of [3].MLP with BP was used. The input layer contains 784 units,
hidden layer contains 400 units and 16 units in the output layer. Three classifiers were proposed.
The first classifier (C1) contains 16-class (Bangla and English numerals) tries to classify the digit,
if the digit was recognized then it further recognizes using 10 class English classifier (C2) else it
uses the 10 class Bangla classifier (C3).7500 Indian postal images were taken. 8690 digits were
used for training which contains (4690 Bangla, 4000 English) and tested with 6406 digits (3179
Bangla and 3227 English).The accuracy achieved for 16-class classifier, Bangla classifier and
English classifier was 92%,94% and 93%.
Velu and Vivekanandan.P [37] have provided a new method for automatic letter sorting for
Indian postal address recognition system based on Pincode. DAB was cropped manually.
Recognition of pin codes was performed using CC approach and ANN. The six nearest neighbour
49
6. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
CC technique was used. The ANN classifier (MLP-CPN) technique was used to recognize the
numerals in the PIN code. ANN was constructed with 64 units in input layer, 100 units in hidden
layer and 52 units in output layer. 6000 samples were used. The experiments were performed on
automatic postal letter sorting machine in AMPS, Chennai. 99.5% of recognition accuracy was
achieved.
Lu et al. [19] have discussed the applications of pattern recognition techniques for postal
automation in China. The analysis was performed on the Siemens sorting machine. The DAB was
extracted using connected component labelling by using Run based algorithm. Features like
mean, standard deviation of merged CC, area, aspect ratio and Euclidean distance. By analysing
the obtained features noise, graphics and characters were classified and by using clustering
algorithm the segments were combined to obtain the AB. The normalized images of the isolated
digits were sent to the classifiers. MLP using BP was performed. Chain Code histograms features
were obtained by dividing the image in 4 x6 grids and by considering 4 direction 96 features were
obtained. Gray scale features were obtained by applying average filters by constructing 8 x12
grids (96), thus 192 features were sent as input and 30 units in the hidden layer and 10 units in the
output layer. Many classifiers like Tree classifiers based on topological features (TCTF), Havlet
and threshold modified Bayesian classifier, Support Vector Machine (SVM) were used and the
result were combined and was chosen based on higher confidence by voting algorithm.
Pfister et al. [27] have developed an OCR-GSA recognition system for handwritten Zip codes in a
real world non- standard letter sorting system. OCR image processing algorithms were used to
read destination address from postal envelopes by Siemens postal automation by Dutch post. Two
classifiers were used to recognize the printed digits. Time delayed neural network (TDNN) was
used to classify the scaled digits feature. While scanning the image the pixels of the image were
grouped in squares called square pixels along with some extracted spatial frequency features.
These were classified using small networks. After calculating the features the square pixels were
clustered and by using some heuristics the address block was located. Zip code was located based
on geometric features like aspect ratio. The TDNN scans the horizontal and vertical pixels using a
pre-set window. NIST data base was used 120000 were trained. And tested on separate set of
images using TDNN.99% accuracy was achieved. The second classifier uses the structural and
quantitative features from the digits pixel. Vectorization was performed by using skeletonization
method. 97% was achieved. When both the classifiers were combined 99% accuracy was
achieved.
Alginaih and Siddiqi [2] have discussed about the multi-stage hybrid Arabic/Indian numeral OCR
system. Three features F1, F2 and F3 were extracted and classified using three classifiers C1, C2
and C3. The first feature F1 was extracted from binary image (array of all pixels), F2 extracted
using zoning technique, by which an array consisting of black pixels were obtained using square
window. F3 was a maximized Fuzzy descriptive feature. The 3 classifiers were C1-Euclidean
distance, C2-Hamming network, C3-Fuzzy neural network. Initially F1 serves as the input for C1
and F2 for C2 and F3 for C3.The output of C1 and C2 were compared if similar then accepted
else again passes through C3.The outputs of C3 is compared with C1 and C2 if matches if any
one classifier then accepted else rejected.400 images were trained and 20 tested.
99.4%
accuracy was achieved.
Hull et al. [11] have proposed a black board based approach for recognition of handwritten Zip
code. Zip code was located using positional information. Three features and three algorithms
were used for classification. The template matching was performed by obtaining the features from
the overall holistic characteristics of the image. A mixed statistical and structural classifier uses
the features from the contours of the digit. Structural classifier uses the information about the size
and displacement of strokes in image. USPS machine printed Zip codes were taken. 1754 digits
50
7. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
were trained and 8129 were tested. The combined classifier performance was tested on 8000
digits and an accuracy of 91% was obtained.
Thomas [35] has developed an automated mail sorter system using SVM classifier. The features
were extracted using Hu moments and SVM classifier was used for recognition. RLSA was used
for image segmentation. The CC analysis was performed based on some features to classify the
text and non-text zones. By using PP histograms were constructed based on the height, width and
density of pixels to locate the Pin code. SVM was used for recognition and 99% accuracy was
obtained. Only isolated Pincode digits were used.
Pervez and Suen [26] have proposed a recognition technique for totally unconstrained
handwritten Zip codes. Recognition methods like statistical and structural methods were used.
The classification module consists of two methods, prediction module which uses the statistical
features and the structural module which uses the structural feature. Prediction module was
performed based on Euclidean distance.5000 digits were trained and 3540 was tested. USPS
database was used. 95% accuracy was achieved. Structural module was computed using Fuzzy
membership. It was trained using 1656 digits and tested using 1103. 99% accuracy was achieved.
When the classifiers were combined and tested the accuracy rose to 96%.
Morshed et al. [24] have performed automatic sorting of mails by recognizing handwritten postal
codes using neural network. DAB and Pincode were located using Heuristics. The neural network
(NN) contains 144 units in input layer, 100 units in the hidden layer and 10 units in the output
layer. 10400 segmented samples from Bangladesh Postal service were used. 5500 samples were
trained and 4900 was tested. 92% accuracy was achieved.
Lu et al. [18] have implemented cost sensitive NN classifiers for post code recognition. MLP with
BP was used. Unequal misclassification costs of all classes were the foundation of cost sensitive
learning. Cost sensitive learning for BP algorithm was used and the following four methods were
experimented cost sampling, cost convergence, rate adapting and threshold moving. 10702
postcode images were taken from Shanghai postal sorting centre. 5351 samples were used for
training and 10702 were tested.81% accuracy was achieved. Their technique can be adapted for
recognition issue of post codes with alphabets or application that have class imbalance problem.
Wang [39] has proposed a recognition system for handwritten Bangla numeral for postal
automation. Principal component analysis (PCA) was used based on image reconstruction,
recognition and direction feature extraction were combined with PCA and SVM was used for
classification. 16000 samples were taken from Bangladesh post. 6000 samples were trained and
10000 samples were tested and 95% accuracy was achieved.
Yonh and et al [42] have presented a new method for recognizing handwritten pin code using
Fuzzy ARTMAP neural network for automatic mail sorting. A special image processing software
WiT was used. Malaysian postal code (five digits) was written on various coloured envelopes
post code box were taken as samples. Region of Interest (ROI) was used to locate the postcode
box. They achieved a recognition accuracy of 90%.
Siva and Uma [32] have proposed a new method to normalize the size of the Zip code image.
Pixel scanning technique was implemented for segmentation and feature extraction of the image.
ANN was used for classification. They achieved 99% accuracy.
Stuti and et al [34] have performed multi script numeral recognition using neural network.
Experiments were performed on five different samples of handwritten Zip codes. They have
experimented with double hidden layer that resulted in 96.53% accuracy. The input layer
51
8. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
constitutes of 150 neurons. The two hidden layers constitute of 250 neurons each and the output
layer is composed of 16 neurons.
Matan et al. [22] have developed a Zip code recognition system for reading handwritten digits.
The main feature of their neural network was that, the input was a pixel image rather than
manually designed feature vectors. Recognition based segmentation was performed. The
segmentation was a hybrid of connected component analysis, vertical cuts and neural network
recognition. Segmentation cuts were performed based on the priori and CCA. The projection
score was used for segmentation based recognition. USPS data base was used. The system was
trained and tested on 10000 images. The accuracy obtained was 87%.
Saifullah and Manry [29] have performed classification based segmentation of Zip codes. A
character classifier was used to determine the trial segmentation that was most likely to be
correct. After segmentation Discrete Fourier Transform (DFT) features were taken. 16 low
frequency DFT features from each segmented character were used. Manual Zip block
segmentation was performed. The segmentation line was chosen based on local maxima and
minima (dips). Many trial dips were performed and based on heuristics the appropriate
segmentation path was chosen.500 Zip codes from USPS database were used. 78 touching Zip
codes using Bayes Gaussian classifier were recognized and the accuracy obtained was 76%.
Wang et al. [38] have used localized arc pattern method for Zip code recognition. Two types of
localized arc patterns, arc (5, 16) and arc (67, 9) were chosen. The first index implies the number
of model patterns and the second index indicates the number of sub areas where frequencies of
the model patterns are counted. The performance of the system highly depends on the threshold
value. 2500 images form the IPTP of Japan mail was used. 90% accuracy was obtained.
Strathy and Suen [33] have proposed a new system for reading handwritten Zip codes.
Segmentation was performed using digit splitter by scanning from left to right in the Zip code
block. The leftmost digit is recognized first and based on that other digits were extracted and
recognized. Digit recognizer uses there NN. Features were extracted using pixel distance feature
method (PDF).BP was used. The networks A, B, and C were constructed based on the following
input, hidden and output units. A and B (336, 70, 10), C (288, 80, 10). The majority vote
combination of 97% was obtained for 2711 images of USPS database.
Lucia Flores [20] has presented a recognition algorithm for printed or handwritten digits of the
Zip code (RAPHZC).The segmentation was performed by scanning the entire digit and copying
all the pixel values into a matrix. The connected digit was segmented based on the quantity of
upper and lower horizontal lines found in the matrix and by using cross points. Four recognition
functions were developed using heuristics. In each of these functions the final matrix was scanned
and by using the crossing points and quantity of pixels the recognition was performed. 22400
handwritten digits and 15120 printed digits were taken from the Brazil correspondences. 99.99%
accuracy was obtained.
Bouchaffra et al. [4] have used non-stationary Markovian models for performing Zip code
recognition. Recognition scores combined with the domain knowledge obtained from the postal
directory files. This data was fed into the model as n-gram statistics that are integrated with
recognition scores of digit images. Gradient structural concavity recognizer was used to recognize
isolated digits. The best model for Zip recognition was chosen based on the experiments that
performed tests on all the possible models in terms of degrees of Markovain process. Their
method merges the images and context in fully Bayesian framework. They obtained promising
results. 20000 Zip codes from USPS database was used and 90% accuracy was obtained.
52
9. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
Mahmoud [21] has performed recognition of digit using Gabor filters. Spatial Gabor filters with
several scales and orientation were used to extract the Gabor features. CENPARM1 (Arabic
Check database) was used. 7390 samples were used for training and 3035 samples were tested.
97% accuracy was obtained for one Nearest Neighbour classifier.
Idan and Chevalier [13] have performed recognition of handwritten digits based on Kohnen’s
self-organizing feature maps. The class labelling of neurons was performed on the map that
emerges during the learning phase. The configuration of the map was set to 8 x 8, 8 x 10 and 11 x
11 and a higher accuracy was 76% was obtained for 11 x 11 topological map. 1000 hand written
Zip code digits were used.735 digits were trained and 265 was tested.
Fujisawa et al. [7] have developed handwritten numeral recognition using gradient and curvature
of gray scale image. Three procedures have been implemented based on the curvature coefficient, bi-quadratic interpolation and gradient vector interpolation was performed for
calculating the curvature of the gray scale curves of an input image. The feature vectors were
obtained using Bayesian approach. Discriminant function was used for classification.12000
images were taken from IPTP CDROM1 of Japan mail was used. The recognition accuracy for
procedures A, B, C was 99.3%, 99.3% and 99.4% respectively. The samples were taken from
Japanese New year greeting card.
Scofield et al. [30] have proposed a multiple neural network architecture (MNNS) or character
recognition. Multiple network architecture was used to combine the response. Low level features
were extracted from local edges, measures of raw pixels image and the image medial axis on the
character contour. NIST database was used.25000 samples were used for training and 4767 was
tested. Four network (MNNS) were constructed and arranged in Hierarchical manner. Restricted
coulomb energy (RCE) was used in 3 layers of feed forward architecture. Cells in the second
layer are radius limited perceptrons referred as Radial Bias function (RBF) used as computer
activation function. Two BP was constructed in the first level and two RCE using RBF was
constructed in the second level. 7300 samples were trained and 2000 samples were tested. The
combined performance was 90%.
4. PERFORMANCE COMPARISON
The following tables display the performance comparison results of all the above mentioned
methods used in the literature. (Table 1) displays the performance comparison of DAB
segmentation. (Table 2) displays the comparison analysis of recognition of Zip code digits using
MLP-BP technique. (Table 3) displays the comparison analysis of recognition of Zip code digits
using other classifiers. The datasets were collected from various countries Postal Service (PS).
Table 1: Performance Analysis of DAB Segmentation
53
10. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
Table 2: Performance Analysis of Zip code Recognition using MLP-BP
54
11. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
(MLP-Multi layer perceptron, BP-Back Propagation, IP-Input layer, HD-Hidden Layer, OPOutput Layer)
55
12. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
Table 3: Performance Analysis of Zip code Recognition using other Classifiers
56
13. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
57
14. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
5. CONCLUSION
The proposed paper explained the various existing methods used in the OCR for automating the
mail sorting process in the literature. It was noted that MLP-BP classifier was widely used for
recognition. The future work mainly concentrates on developing a fast and cost effective
algorithm for extraction of DAB and recognition Pin code for the business mail service of Indian
post.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
E. Akira Yonekura and J. Facon. (2003),“2-D histogram-based segmentation of postal
envelopes”, In Proceedings of 16th SIBGRAPI Brazilian Symposium on Computer Graphics and
Image Processing, pp: 247-253.
Yasser M. Alginaih and Abdul Ahad Siddiqi. (2010),“Multistage hybrid Arabic/Indian
numeral
OCR system”, International Journal of Computer Science and Information Security.
Gerson Bochnia and Jacques Facon.“Address block extraction of magazine envelopes by multiple
threshold”. Rua Imaculada Conceic¸ao˜ 1155, Prado Velho, 80215-901 Curitiba-Pr, Brazil.
Djamel Bouchaffra, Sargur Srihari and Venu Govindaraju. (1999).“Postprocessing of recognized
strings using nonstationary Markovian models”. IEEE Transactions on Pattern Analysis and
Machine Intelligence, Vol. 21, pp: 990-999.
Luiz Felipe Eiterer, Jacques Facon, and David Menoti. (2004),“Postal envelope address block
location by fractal-based approach”, In Proceedings of 17th Brazilian Symposium on Computer
Graphics and Image Processing, pp: 90-97.
Jacques Facon, David Menoti, and Arnaldo de Albuquerque Araujo. (2005),“Lacunarity as a texture
measure for address block segmentation”. In Proceedings of the 10th IberoAmerican Congress
conference on Progress in Pattern Recognition, Image Analysis and Applications, pp: 112-119.
Y. Fujisawa, Meng Shi, T.Wakabayashi, and F. Kimura. (1999), “ Handwritten numeral recognition
using gradient and curvature of gray scale image”. In Proceedings of the 5th International
Conference on Document Analysis and Recognition, ICDAR, pp: 277-280.
Djamel Gaceb, Veronique Eglin, Frank Lebourgeois, and Hubert Emptoz. (2009), “Robust approach
of address block localization in business mail by graph coloring”, The International Arab Journal of
Information Technology.
Xue Gao and Lianwen Jin. (2009),“Swift post a vision-based fast postal envelope identification
system”. In Proceedings of the 2009 IEEE International Conference on Systems, Man, and
Cybernetics, pp: 3268-3273.
Venu Govindaraju and Sergey Tulyakov. (2003), “Postal address block location by contour
clustering”. In Proceedings of the Seventh International Conference on Document Analysis and
Recognition, Vol.1, pp: 429.
Jonathan J. Hull, Sargur N. Srihari, Leonard Kuan, Peter Cullen, and Paul Palumbo. (1988), “A
blackboard-based approach to handwritten Zip code recognition”. In Proceedings of 3rd United
States Postal Service Advanced Technology Conj, pp:1018-1032.
Chabaan Ibrahim. (2007), “Applying Domain Knowledge To The Recognition Of Handwritten Zip
Codes”. PhD thesis, Indiana University, South Bend, US.
Y. Idan and R.C. Chevalier. (1991), “Handwritten digits recognition by a supervised Kohonen-like
learning algorithm”. In IEEE International Joint Conference on Neural Networks, Vol. 3, pp: 25762581.
Mustafa Idrissi. (2008),“Address block segmentation using ensemble clustering techniques”, PhD
thesis, Faculty of Electrical Engineering, Mathematics and Computer Science, Delft,
F. Kimura, Y. Miyake, and M. Shridhar. (1995),“Handwritten Zip code recognition using lexicon
free word recognition algorithm”. In Proceedings of the 3rd International Conference on Document
Analysis and Recognition, Vol. 2, pp: 906.
Y. Le Cun, O. Matan, B. Boser, D. Denker, J.S. Henderson, W.Jacket, L.D.Baird, H.S. Howard, R.E.
Hubbard. (1990), “Handwritten Zip code recognition with multilayer networks”. In Proceedings of
the 10th International Conference on Pattern Recognition.
58
15. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
Horacio Andres Legal-Ayala and Jacques Facon. (2003), “Image segmentation by learning
approach”. In Proceedings of the 7th International Conference on Document Analysis and
Recognition, Vol.2, pp: 819.
Shujing Lu, Li Liu, Yue Lu, and Patrick S. P. Wang. (2012), “ Cost-sensitive neural network
classifiers for postcode recognition”. International Journal of Pattern Recognition and Artificial
Intelligence, Vol.26, No.5.
Yue Lu, Xiao Tu, Shujing Lu, and Patrick S P Wang. (2010), “Application of pattern recognition
technology to postal automation in china”. Pattern Recognition and Machine Vision.
E. Lucia Flores. (1998), “A recognition algorithm of printed or handwritten digits of the Zip code”.
In proceedings of Telecommunications Symposium, IEEE International, vol. 2, pp: 671-676.
S.A. Mahmoud. (2009),“ Recognition of Arabic (Indian) check digits using spatial Gabor Filters”. In
GCC Conference Exhibition, IEEE, pp: 1-5.
Ofer Matan, Henry S. Baird, Jane Bromley, Christopher J. C. Burges, John S.Denker, Lawrence D.
Jackel, Yann Le Cun, Edwin P. D. Pednault, William D.Satterfield, Charles E. Stenard, and Timothy
J. Thompson. (1992), “Reading handwritten digits: A Zip code recognition system”. Computer,
Vol.25, No.7, pp:59-63.
David Menoti, Dibio Leandro Borges, Jacques Facon, and Alceu de Souza Britto, Jr. (2003),
“Segmentation of postal envelopes for address block location: an approach based on feature selection
in wavelet space”. In Proceedings of the 7th International Conference on Document Analysis and
Recognition, Vol.2, pp: 699.
A.S.M.Mahbub Morshed, Md.Shahidul Islam, Md.Abul Faisal, Md.Abdus Sattar, and Mitsuru
Ishizuka. (2007), “Automatic sorting of mails by recognizing handwritten postal codes using neural
network architectures”. The Pennsylvania State University.
Umapada Pal, Rami Kumar Roy, Kaushik Roy, and Fumitaka Kimura. (2009), “Indian multi-script
full pin-code string recognition for postal automation”. In Proceedings of the 10th International
Conference on Document Analysis and Recognition, pp: 456-460.
Pervez and C.Y. Suen. (1986), “Computer recognition of totally unconstrained handwritten Zip
codes”. International Journal of Pattern Recognition and Artificial Intelligence.
M. Pfister, S. Behnke, and R. Rojas. (2000), “Recognition of handwritten Zip codes in a real world
non-standard letter sorting system”. Applied Intelligence.
K. Roy, S. Vajda, U. Pal, and B. B. Chaudhuri. (2004), “ A system towards Indian postal
automation”. In Proceedings of the Ninth International Workshop on Frontiers in Handwriting
Recognition, pp: 580-585.
Y. Saifullah and M.T. Manry. (1993), “Classification-based segmentation of Zip codes”. IEEE
Transactions on Systems, Man and Cybernetics,Vol.23, No.5, pp:1437-1443.
C.L. Scofield, L. Kenton, and J.C. Chang. (1991), “Multiple neural net architectures for character
recognition”, In Compcon. Spring. Digest of Papers, pp: 487-491.
Prasun Sinha. (1997), “Cursive Script Postal Address Recognition”. PhD thesis, Michigan State
University, East Lansing, US.
Kumar.K. Siva and Devi.D Uma. (2012), “Handwritten Zip code classification using artificial neural
networks”, International Journal of Engineering Research Technology, Vol.1, No.6.
N. W. Strathy and C. Y. Suen. (1995), “A new system for reading handwritten Zip codes”, In
Proceedings of the 3rd International Conference on Document Analysis and Recognition, Vol.1, pp:
74.
Asthana Stuti, Farha Haneef and R.K Bhujade. (2011), “Handwritten multi-script numeral
recognition using artificial neural networks”, International Journal of Soft Computing and
Engineering, Vol.1, No.1.
Arun K.S.Jerin Thomas. (2011), “An automated mail sorter system using SVM
classifier”,International Journal of Computer Applications.
S. Vajda, K. Roy, U. Pal, B. B. Chaudhuri, and A. Belaid. (2009), “Automation of Indian postal
documents written in Bangla and English”. International Journal of Pattern Recognition and
Artificial Intelligence, Vol.23, No.8, pp:1599-1632.
C. M. Velu and Vivekanandan.P. (2010), “Automatic letter sorting for Indian postal address
recognition system based on pin codes”, Journal of internet and information system,Vol.1, No.6,
pp:15.
XianWang, Venu Govindaraju, and Sargur Srihari. (1999), “Multi-experts for touching digit string
recognition”. In Proceedings of the 5th International Conference on Document Analysis and
Recognition, pp: 800.
59
16. Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013
[39]
[40]
[41]
[42]
[43]
Yi-Kai Chen Jhing-FaWang. (2001), “Locating the destination address block on images of complex
mail pieces”, Journal of the Chinese Institute of Engineers.
Ralph Wolf and John C. Platt. (1994), “Postal address block location using a convolutional locator
network”, Advances in neural information processing system,Vol.6,pp:745-752.
Junliang Xue, Xiaoqing Ding, Changsong Liu, Rui Zhang, and Weiwei Qian. (2001), “Location and
interpretation of destination addresses on handwritten Chinese envelopes”, Pattern Recognition.
Letters, Vol.22, No.6-7, pp:639-656.
H.T Yonh, K.Marzuki, T.K.Kok and Y.Rubyiah. (1997), “Hand-written post code recognition by
Fuzzy Artmap neural network”. In Proceedings of the Confederation of Scientific and Technological
Association.
Bin Yu, A.K. Jain, and M. Mohiuddin. (1997), “Address block location on complex mail pieces”. In
Proceedings of the 4th International Conference on Document Analysis and Recognition, Vol. 2, pp:
897-901.
AUTHORS
Dr.R.Radha, aged 44, working as assistant Professor in Shrimathi Devkunvar Nanalal
Bhatt Vaishnav College for Women, Chromepet, Chennai, Tamil Nadu, India. She has
20 years of teaching experience. She did her research in Fuzzy Based Data Mining For
Effective Decision Support In Bio-Medical Applications. Her research interest is in bio
informatics and image processing. She has published 15 papers.
R.R.Aparna a research scholar in Madras university. Her research interest includes
Image processing and Artificial Neural networks.She has published 2 papers.
60