I. INTRODUCTION:

Image Quality Assessment (IQA) has always been an integral part of image processing. Many different approaches for IQA with different density have been developed in the last decade. Digital images are subject to a variety of distortions during compression, transmission, processing, and reproduction. In order to maintain, control and possibly enhance the quality of the image and video data being delivered, it is important for data management system (network video servers) to be able to identify and quantify quality degradations on the fly. [1] Image QA methods can be classified as subjective and objective methods. The first approaches to image quality evaluation are subjective quality testing which is based on observers that evaluate image quality. These tests are time consuming, expensive and have a very strict definition of observational conditions. The second approaches are the objective image quality testing based on mathematical calculations. Over the years, a number of researchers have contributed significant research in the design of full reference image quality assessment algorithms, claiming to have made headway in their respective domains. The QA research community realizes the importance of validating the performance of algorithms using extensive ground truth data, particularly against the backdrop of the fact that a recent validation study conducted by the video quality experts group (VQEG) discovered that the nine video QA methods that it

tested, which contained some of the most sophisticated algorithms at that time, were “statistically indistinguishable” from the simple peak-signal-to-noise-ratio (PSNR) [2].

It is therefore imperative that QA algorithms be tested on extensive ground truth data if they are to become widely accepted. Furthermore, if this ground truth data, apart from being extensive in nature, is also publicly available, then other researchers can report their results on it for comparative analysis in the future.

In this paper we present our results of a wide subjective quality assessment study, and estimate the concert of six recognized QA algorithms. The psychometric study contained 200 images distorted using different distortion types and more than 500 human image quality evaluations. This study was miscellaneous in terms of image content, distortion types, distortion strength, as well as the number of human subjects ranking each image. We have also made the data set publicly available [3] to facilitate future research in image quality assessment.

In the current connected world, many users share and deliver multimedia data. The overall communication process includes manipulation, processing, storing, and transmission over (noisy) channels. Although there have been great improvements in compression and transmission techniques, each stage of processing may introduce perceivable distortions [4,5]. For example, blocking, ringing, and blurriness are only few of the artifacts that a lossy compression algorithm introduces in an image.

II. METHODOLOGY:

Step 1

Figure 1: Process of efficiency verification for a given method of image filtering.

As a first step, a trial image (or a set of test images) presenting good value is selected. Then, according to the selected model of noise or distortion, a noisy version of the picture (images) is obtained and processed by a planned filter. The obtained output image is”compared” to the matching original image using measured quality metric. A value of the same metric is calculated for the noisy (distorted) image as well. By comparing the scores of these metrics it is possible to address the effectiveness of the designed filtering technique.

Step-2

Figure 2: Process of efficiency verification for a given lossy compression technique.

In second step, a quality metric can be used both in the design of the compression block and in the overall performances evaluation. In the concluding case, a metric value calculated for a decoded image can be used in the tuning phase of the parameters in the coarse-to-fine compression schemes. Then if, for example, an obtained value of quality metric is unsuitable, an image can be compressed with better quality with smaller quantization step or, equivalently, with larger bit rate.

Digital image processing techniques involves a variety of methods such as image filtering, reconstruction, inpainting, etc. For this class, image visual quality metrics are used only in the process of a method design and estimation of its efficiency.

Quality Assessment Parameters (QAP):

1. Mean Square Error (MSE):

MSE measures the average of the square of the "error." The error is the amount by which the estimator differs from the quantity to be estimated. The difference occurs because of randomness or because the estimator doesn't account for information that could produce a more accurate estimate.

2. Peak Signal to Noise Ratio (PSNR):

The phrase peak signal-to-noise ratio, often abbreviated PSNR, is an engineering term for the ratio between the maximum possible power of a signal and the power of corrupting noise that affects the fidelity of its representation.

3. Structural Content (SC):

The loss in perceived image quality is often determined by the nature and level of an artifact along with the context in which it appears. For example, in a highly structured image containing lines and edges, sharpness will likely be the most critical attribute in ranking image quality; whereas, low-frequency uniformity may have little impact on the quality decision

4. Average Difference (AD):

The difference in visual properties that makes an object (or its representation in an image) distinguishable from other objects and the background. In visual perception of the real world, contrast is determined by the difference in the color and brightness of the object and other objects within the same field of view. Because the human visual system is more sensitive to contrast than absolute luminance, we can perceive the world similarly regardless of the huge changes in illumination over the day or from place to place.

5. Normalized Absolute Error (NAE):

Normalization is the process of isolating statistical error in repeated measured data. Normalization is sometimes based on a property. Quintile normalization, for instance, is normalization based on the magnitude (quintile) of the measures.

In another usage in statistics, normalization refers to the division of multiple sets of data by a common variable in order to negate that variable's effect on the data, thus allowing underlying characteristics of the data sets to be compared: this allows data on different scales to be compared, by bringing them to a common scale. In terms of levels of measurement, these ratios only make sense for ratio measurements

III. RESULT ANALYSIS:

(a) Original Image

(b)

(b) Distorted_1 (c) Distorted_2

(d) Distorted_3 (e) Distorted_4

Figure 3: Comparing with Motion Blurred Image

Parameters	MSE	PSNR	AD	NAE	SC
Original	0	99	0	0	1
Distorted_1	152.329	26.303	-1.95729	0.7002	0.9912
Distorted_2	126.621	27.105	-1.05356	0.0560	0.9871
Distorted_3	132.861	26.898	-2.02498	.07742	0.9613
Distorted_4	105.861	27.883	-1.95084	0.0605	0.9740

(a) Original Image

(b) Distorted_1 (c) Distorted_2

(d) Distorted_3 (e) Distorted_4

Figure 4: Comparing with Distorted Image

Parameters	MSE	PSNR	AD	NAE	SC
Original	0	99	0	0	1
Distorted_1	1006.6	18.10	-17.8	0.30	0.80
Distorted_2	2221.5	24.67	-2.80	0.95	0.97
Distorted_3	1221.1	17.26	-29.9	0.39	0.70
Distorted_4	8998.1	8.58	73.50	0.95	369.8

IV. CONCLUSION:

The results show large sensitivity variations among the different methods. Most of the algorithms implemented here have been extended to evaluate the quality of images. The Peak Signal Noise Ratio (PSNR) is higher since in HSV color space there is normalization of pixel values so Mean Square Error (MSE) is very small, so little chance for comparison. It is clearly observed that the Structural Content (SC) is considerably enhanced, but still PSNR is beneficial for getting better quality when dealing with compressed images.

V. FUTURE SCOPE OF THE WORK:

No reference Image Quality Assessment provides the way, in which we evaluate the quality of an Image without having a referenced image. It is applicable in those areas where a copy of single image is sent via mail or though any other media and the quality evaluation are needed.

VI. REFERENCES:

1. Zhou Wang, Member, IEEE, Guixing Wu, Hamid Rahim Sheikh, Member, IEEE, Eero P. Simoncelli, Senior Member, IEEE, En-Hui Yang, Senior Member, IEEE, and Alan Conrad Bovik, Fellow, IEEE, “Quality Aware Images”, IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 6, JUNE 2006.

2. VQEG, “Final report from the video quality experts group on the validation of objective models of video quality assessment,” http://www.vqeg.org/ , Mar. 2000.

3. A. M. Eskicioglu and P. S. Fisher, “Image quality measures and their performance,” IEEE Trans. Communications, vol. 43, no. 12, pp. 2959–2965, Dec. 1995.

4. M. Yuen and H. R. Wu, "A survey of hybrid MC/DPCM/DCT video coding distortions," Signal Processing, vol. 70, pp. 247-78, 1998.

5. A. B. Watson, “Digital Images and Human Vision”, MIT Press, ch.3, pp. 139-140, 1993.

Received on 20.08.2011 Accepted on 05.10.2011

Int. J. Tech. 1(2): July-Dec. 2011; Page 68-71