Structural Similarity Based Image Quality Assessment Using Full Reference Method

This paper presents an objective quality assessment for digital images that have been degraded by noise. Objective quality assessment is crucial and is generally used in image processing. The main objective of this paper is to analyse various statistical properties and their measurements and finally compare them. The statistical properties that are included are mean square error (MSE), ischolar_main mean square error (RMSE), signal to noise ratio (SNRQ), peak signal to noise ratio (PSNR) and certain frequency parameters like spectral magnitude distortions and spectral phase distortions. But it is observed that MSE and PSNR yield poor results therefore a new metric namely structure similarity is proposed which has a better performance than MSE and PSNR but fails when applied on badly blurred images. Therefore, edge based structure similarity index metric (ESSIM) is proposed. Experiment results show that ESSIM is more consistent with human visual system (HVS) than SSIM and PSNR especially for the blurred images.