1. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
FULL- REFERENCE METRIC FOR IMAGE QUALITY ASSESSMENT
Ms. Shraddha N. Utane Prof. V. K. Shandilya
Department Of Information Technology Department Of Information Technology
Sipna’s COET(Amravati University) Sipna’s COET(Amravati University)
Amravati, India Amravati , India
Abstract-The quality of image is most important factor in [10] can be categorized into subjective and objective
image processing, to evaluate the quality of image various methods.The former is based on the quality which is
methods have been used. Proposed system defines one of the assessed by human observers, and the latter provides an
best methods in image quality assessment. Proposed system objective index or real value which is obtained from any an
calculates the image quality assessment using normalized
histogram. Sender send the image to the receiver, after
assessment model to measure the image quality. Because
receiving the image, receiver compare the image with senders human observers are the ultimate receivers of the visual
image using normalized histogram. In proposed work, MGA information contained in an image, subjective method [5]
transforms perform excellently for reference image whose results are directly given by human observers is
reconstruction, have perfect perception of orientation, are probably a reliable way to assess the quality of an image.
computationally tractable, and are sparse and effective for The subjective method is that the observers are asked to
image representation. MGA is utilized to decompose images evaluate the picture quality of sequences using a continuous
and then extract features to mimic the multichannel structure grading scale and to give one score for each sequence. A
of HVS. Additionally, MGA offers a series of transforms number of different subjective methods are represented by
including wavelet, curvelet, bandelet, contourlet transform etc.
These different types of transforms are used to capture
ITUR Recommendation BT.500[4]. The subjective quality
different types of image geometric information. Contrast measurement has been used for many years as the most
Sensitivity Function(CSF) and Just Noticiable Difference(JND) reliable form of quality measurement. However, subjective
isused to produce a noticeable variation in sensory experience. experiment requires human viewers working in a long
After calculating the normalized histogram of both the period, and repeated experiments are needed for many
reference and distorted image, we are checking the quality of image objects, it is expensive and time consuming, and
both the images. cannot be easily and routinely performed for many
scenarios, e.g., real time systems. Moreover, there has not
Index Terms- Multiscale Geometric Analysis(MGA), Full- been any precise mathematical model for subjective
reference(FR) metric, Human Visual System(HVS).
assessment currently. As a consequence, there arouses the
requirement of an objective quality metric that accurately
I. INTRODUCTION
matches the subjective quality and can be easily
The objective of image quality assessment (IQA) [1] is to
implemented in various image systems, leading to the
provide computational models to measure the perceptual
emergence of the objective IQA. The objective IQA [3] is
quality of an image. In recent years, a large number of
proposed to provide a computational model to measure the
methods have been designed to evaluate the quality of an
perceptual quality of an image. It makes use of the variation
image, which may be distorted during acquisition,
of several original or distorted image characteristics which
transmission, compression, restoration, and processing
is caused by degradation to represent the variation of the
which lead to image degradation. In poor transmission
image perceptual quality. Many objective quality metrics for
channels, transmission errors or data dropping would
predicting image distortions have been investigated. Metrics
happened, which lead to the imperfect quality and distortion
are usually obtained from either reference or distorted
of the received video data. Therefore, how to evaluate the
images to reflect a number of image characteristics.
image quality [2]has become a burning problem. In recent
Evaluation results obtained from a good objective IQA
years, digital camera is equipped in most of the mobile
method should be statistically consistent with subjective
products like cellular phone, PDA and notebook computer.
methods. According to the availability of a reference image,
Image quality is the most important criteria to choose
there is a general agreement [1] that objective quality
mobile products. In some cases, the benchmarks or reviews
metrics can be divided into three categories: full-reference
of products are based on subjective image quality test and
(FR), no-reference (NR), and reduced-reference (RR)
thus are dependent on tester and environment. The
method.
subjective image quality assessment often misleads the
decision for the image quality control parameters of Image
II. FULL- REFERENCE METRIC (FR)
Signal Processing (ISP) algorithm.The recent years has
demonstrated and witnessed the tremendous and imminent
In this metric we are sending the reference or original image
demands for image quality assessment (IQA) methods at
to the receiver side. Then with the help of quality evaluation
least in the following three ways: 1) They can be exploited
system we are comparing both the original and degraded
to monitor image quality for controlling quality of
image. Evaluation system consist of two metrics within this
processing system. 2) They can be employed to benchmark
class are the PSNR (Peak Signal-to-Noise Ratio) and the
image processing systems and algorithms. 3) They can also
MSE (mean square error), due to simplicity of their
be embedded into image processing systems to optimize
computation. Conventional FR IQA methods calculate
algorithms and parameter settings. Existing IQA metrics
pixel-wise distances, e.g., peak signal-to-noise ratio(PSNR)
149
All Rights Reserved © 2012 IJARCSEE

2. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
and mean square error (MSE), between a distorted image the system is depicted in figure 2. As an example, Webster et al.
and the corresponding reference, but they have not been in propose a reduced reference system in which the side information
agreement with perceived quality measurement widely [1]. consists of two distinct types of measurements: spatial
measurements extracted from the frames edges, and temporal
measurements extracted from frames differences.
Fig.1.1 Full reference metrics
III. NO-REFERENCE METRIC (NR) :
The evaluation system has no reference to any side information Fig 1.3 Reduced-reference metric
regarding the original media. This kind of metrics is the most
promising in the context of video broadcast scenario, since the
original images or video are in practice not accessible to end users.
V. PROPOSED WORK
In proposed framework implementation of image quality
assessment will be carried out by considering various types
of images from LIVE database. In this we are considering
the reference images and distorted images distorted by some
parameter. The working of proposed framework is listed
below.
Module 1: Image Decomposition for Feature Extraction.
In which we will decompose the image using MGA
framework for feature extraction. It [5] is popular to analyze
Fig 1.2 No-reference metric signals in both time and frequency domains simultaneously
and adaptively. By using multiscale operation, it extracts
In the no-reference objective metrics scenario (figure 2), quality effective features to represent signals, especially for
rating is attained through analysis of the received media only. No- nonstationary signals . In our proposed work, we are using
reference objective metrics are relatively rare in literature, but
bior wavelet transform to decompose the image into
some proposals have been made. Generally, the proposed
algorithms evaluate some specific quality features that result from
highpass subbands and a lowpass residual subband. Both
image or video transmission, like block effect in block-based DCT the reference and distorted images are decomposed by using
compression methods, edge discontinuity, etc. This kind of wavelet transform.
analysis is possible by taking into account both human visual Module 2: CSF Masking
system models and natural image models. By contrast, designing In second module after decomposition of the images we are
objective No-Reference (NR) quality measurement algorithms is a applying the CSF masking .CSF [6] measures how sensitive
very difficult task. This is mainly due to the limited understanding we are to the various frequencies of visual stimuli, i.e., we
of the HVS, and it is believed that effective NR quality assessment are unable to recognize a stimuli pattern if its frequency of
is feasible only when the prior knowledge about the image
visual stimuli is too high. It Re-weights MGA decomposed
distortion types is available [8].
coefficients to mimic the nonlinearities inherent in HVS. It
IV. REDUCED-REFERENCE METRIC (RR): measures sensitivity to various frequency of visual stimuli.
Module 3: JND Threshold
After applying the CSF masking we are calculating the
The evaluation system has access to a small amount of side threshold value.It measures the minimum amount, by which
information regarding the original media, i.e. features or stimulus intensity must be changed to produces a noticeable
descriptors extracted from the original. In a reduced reference variation in sensory experience. Lower the JND threshold,
metrics scenario, the content provider transmits additional visual quality of reconstructed image is better.
information together with the video. This class of metrics requires Module 4: Normalized Histogram for Image
additional bandwidth (or an additional channel) to transmit the side Representation
information. Generically speaking, side information usually Histogram is constructed for both the reference image and
consists of relevant features extracted from the original media the distorted image and both images are compared for
which are transmitted and compared with the analogous features
quality.
extracted from the degraded media. The amount of additional
information that is transmitted through the side channel is highly Module 5: Evaluation of Quality of Image
dependent of the design of the system. The general architecture of
150
All Rights Reserved © 2012 IJARCSEE

3. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
We will evaluate the quality of image using proposed
framework. In this we will evaluate the quality of image by
using different metric such as SSIM ,FSIM etc.
Input Image
Evaluation Of
.
Quality Of
Image
Reference Distorted
image image
FSIM SSIM WSNR
Image Image
Decompositi Decompositi
on Using on using
Wavelet Wavelet
Fig.3 Evaluation Metric For Image Quality
transform transform
Applying the Applying the
CSF CSF VI. STRUCTURAL SIMILARITY INDEX METRIC
Masking and Masking and
calculating calculating The structural similarity (SSIM) [9]index is a method
the the for measuring the similarity between two images. The
SSIM index is a full reference metric, in other words,
Threshold Threshold the measuring of image quality based on an initial
value value uncompressed or distortion-free image as reference.
SSIM is designed to improve on traditional methods
like peak signal-to-noise ratio (PSNR) and mean
squared error (MSE), which have proved to be
. Construct Construct inconsistent with human eye perception. The difference
Normalized Normalized with respect to other techniques mentioned previously
Histogram Histogram such as MSE or PSNR, is that these approaches
estimate perceived errors on the other hand SSIM
considers image degradation as perceived change in
structural information. Structural information is the idea
that the pixels have strong inter-dependencies especially
Calculate Calculate when they are spatially close. These dependencies carry
Quality of Quality of important information about the structure of the objects
in the visual scene.The SSIM metric is calculated on
Image image various windows of an image. The measure between
two windows and of common size N×N is:
Evaluation
VII. FEATURE SIMILARITY INDEX METRIC
of Quality
Of
of image Image quality assessment (IQA) aims to use computational
models to measure the image quality consistently with
Fig.2 Framework of the proposed model subjective evaluations. The well-known structural-similarity
(SSIM) index brings IQA from pixel based stage to structure
based stage. In this work, a novel feature-similarity (FSIM)
151
All Rights Reserved © 2012 IJARCSEE

4. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
[8]index for full reference IQA is proposed based on the fact framework has better effectiveness not only than the
that human visual system (HVS) understands an image reduced-reference method, WNSIM, but also than the full
mainly according to its low-level features. Specifically, the reference method as MSSIM, in terms of CC, ROCC, OR,
phase congruency (PC), which is a dimensionless measure MAE, and RMSE. In addition, different transforms have
of the significance of a local structure, is used as the primary different advantages in different distortions. For example,
feature in FSIM. Considering that PC is contrast invariant wavelet transform is optimal to Gblur distortion, curvelet
while the contrast information does affect HVS’ perception transform and PSNR perform best onWNdistortion, and
of image quality, the image gradient magnitude (GM) is WBCT outperforms the other methods for FF distortion.
employed as the secondary feature in FSIM. PC and GM 3) Low data rate: The proposed framework has a relatively
play complementary roles in characterizing the image local low data rate for representing features of the reference
quality. After obtaining the local similarity map, we use PC image, i.e., a relatively low bits are used to represent the
again as a weighting function to derive a single quality image features (e.g., for wavelet-based decomposition,there
score. Extensive experiments performed on six benchmark are only 10 features/image utilized for
IQA databases demonstrate that FSIM can achieve much representation;contourlet-based decomposition corresponds
higher consistency with the subjective evaluations than all to 17 features/image; WBCT-based decomposition
the state-of-the-art IQA metrics used in comparison. corresponds to 25 features/image; and HWD-based
Although FSIM is designed for grayscale images (or the decomposition corresponds to 17 features/image).
luminance components of color images), the chrominance XI. CONCLUSION
information can be easily incorporated by means of a simple
extension of FSIM, and we call this extension FSIMC. In this proposed framework , a Full-reference image quality
assessment framework is proposed by incorporating merits
VIII. WEIGHTED SIGNAL TO NOISE RATIO (WSNR) of multiscale geometry analysis (MGA), contrast sensitivity
function (CSF), and the Weber’s law of just noticeable
In [10], a different approach to PSNR was presented: As the difference (JND). In comparing with existing image quality
human visual system (HVS) is not equally sensitive to all assessment approaches, the proposed one has strong links
spatial frequencies, a contrast sensitivity function (CSF) is with the human visual system (HVS): sparse image
taken into account. The CSF is simulated by a lowpass or representation is utilized to mimic the multichannel structure
bandpass frequency filter.First of all, the difference of the of HVS, CSF is utilized to balance magnitude of coefficients
reference and the distorted image is computed. Then the obtained by MGA to mimic nonlinearities of HVS, and JND
difference is transformed into frequency domain using 2- is utilized to produce a noticeable variation in sensory
dimensional fast Fourier transform. The obtained error experience. In this framework, images are represented by
spectrum is weighted by the CSF resulting in weighted error normalized histograms, which correspond to visually
spectrum. The last thing to do is to compute the power of the sensitive coefficients. The quality of a distorted image is
weighted error spectrum and the power of the signal (also measured by comparing the normalized histogram of the
transformed into frequency domain). distorted image and that of the reference image.
IX. TABLE
Metric Reference Image Distorted Image
XII. ACKNOWLEDGEMENT
FSIM 0.8611 0.8567
I express my sincere thanks to Prof.V.K.Shandilya for her
PraposedFramework 77.9088 33.5261 valuable guidance.
Table 1. Quality Evaluation Parameter
X. ADVANTAGES XIII. REFERENCES
1) General purpose: A number of different transforms, e.g., [1]. Z. Wang and A. C. Bovik, Modern Image Quality
wavelet, curvelet, bandelet, contourlet, WBCT and Assessment. New York: Morgan & Claypool, 2006.
HWD,for image decomposition can be applied in the .[2] Wang, Z., Bovik, C. A., and Lu, L. G., "Why is image
proposed framework for IQA. All these transforms can work quality assessment so difficult?" in Proc. IEEE Int. Conf.
well for different image distortions, and WBCT and HWD Acoustics, Speech, and Signal Processing, Florida, USA, 4,
perform much better than the others, especially for JPEG 3313-3316 (2002).
and JPEG2000 images. [3] Wang, Z. and Bovik, C. A., Modern Image Quality
2) Sound effectiveness: The objective assessment of the Assessment, New York: Morgan and Claypool Publishing
proposed framework performs consistently with the Company (2006).
subjective perception, and can evaluate visual quality of [4] Methodology for the Subjective Assessment of the
JPEG and JPEG2000 image effectively. Particularly, by Quality of Television Pictures, Recommendation ITU-R
applying WBCT and HWD for image decomposition, the Rec. BT. 500-11.
152
All Rights Reserved © 2012 IJARCSEE

5. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
[5] S. Gabarda and G. Cristobal, “Blind image quality [8] FSIM: A Feature Similarity Index for Image Quality
assessment through anisotropy,” J. Opt. Soc. Amer. A, vol. Assessment Lin Zhanga, Student Member, IEEE, Lei
24, pp. B42–B51, 2007 Zhanga,1, Member, IEEE Xuanqin Moub, Member, IEEE,
[6] M. Miloslavski and Y.-S. Ho, “Zerotree wavelet image and David Zhanga, Fellow, IEEE
coding based on the human visual system model,” in Proc. [9] Video Quality Metrics Mylène C. Q. Farias
IEEE Asia-Pacific Conf.Circuits and Systems, 1998, pp. 57– Department of Computer Science University of Brasília
60 (UnB) Brazil.
[7] S. Mallat, “A theory for multiresolution decomposition: [10] Damera-Venkata, N. et al. Image Quality Assessment
The wavelet representation,” IEEE Trans. Pattern Anal. BasedonDegradationModelhttp://www.ece.utexas.edu/~bev
Mach. Intell., vol. 11, no. 7, pp. 674–693, Jul. 1989 ans/papers/2000/imageQuality/
153
All Rights Reserved © 2012 IJARCSEE