Image quality is a measure of perceived naturalness and usefulness . Since image quality plays a crucial role in image processing based applications, it is necessary to quantify the amount of visual distortion in images introduced during various processes like acquisition, compression and transmission.
Image quality assessment (IQA) attempts to quantify visual quality in terms of metrics, which are used for benchmarking and comparing performances of compression techniques. Basically, Image Quality assessment is done by Objective and Subjective approaches. An Objective Image quality assessment is further categorized into 1) Full Reference (FR) 2) No Reference (NR) and 3) Reduced Reference (RR). In the benchmarking of the image compression and restoration algorithms, an original image becomes an ultimate reference for evaluation of the processed image. Quality assessment done by comparing the processed image with original image of high quality called as full reference (FR) image quality assessment. No reference (NR) image quality assessment is carried out when the reference image is not available at the time of assessment, while in the reduced reference (RR) image quality assessment some extracted features of reference image describing the image information are made available at the time of assessment .
Conventionally, image quality is measured by subjective approach where a set of images is shown to group of trained people and they are asked to record their opinion about image quality. Mean of all such ratings is called as Mean Opinion Score (MOS). If the reference image is also shown with the distorted image then Differential Mean Opinion Score (DMOS) is obtained as the difference between MOS of reference image and test image. This DMOS value is zero for original image and increases as the amount of distortion increases.