S-CIEL *a*b* Based Quality Assessment of Images Restored from Non-Gaussian-Noise-Corrupted-Images and Non-Linear-Bilateral-Kernel Based Restoration

Most widely used image-quality-assessment- metrics viz., Cumulative Minimum-Mean-Square-Error (CMMSE) and Cumulative Peak-Signal-to-noise-ratio (CPSNR) do not satisfactorily measure the quality of image-restoration. As it is an observed empirical fact amongst the fraternity of researchers that a difference of ten or even twenty percentage in SNR may not at all translate into any difference in the perceivable image quality. And, by some construction, images that have similar PSNR but of drastically different attributes can conceivably created. This challenge is addressed empirically in this study by using S-CIEL*a*b* image-assessment-quality-metrics. Of course, spatially extended CIEL*a*b* (S-CIEL*a*b*) is primarily used to represent colors numerically and to calculate the color differences for the scholarly reason that the human vision can perceive millions of color differences; but no technology can meaningfully replicate the ability of human eye as closer as the CIEL*a*b* can perform. Another unique feature of the research study is that the restoration is achieved from the non-Gaussian-noise-corrupted-natural images. It is an observed fact that linear filters blur the restored image; here in this investigative study the non-linear-bilateral filter technique is implemented in order to establish the superior performance of the said assumptions and objectives of the study. The standard datasets are used for carrying the experimentation.