Video quality assessment based on a modified mean squared error

Abstract

Video quality is defined as how much the similarity of the video to its original undistorted video. In order to judge the video quality without human interception, video quality assessment (VQA) methods are proposed. Most of the VQA were developed from the popular image quality assessment (IQA) methods such as mean squared error (MSE) or peak signal to noise ratio (PSNR), structural similarity (SSIM) and multi-scale SSIM (MSSIM). However, different from the theory of IQA, VQA involves temporal distortions and effects. Due to the temporal effect, videos are affected by more distortions or errors as compared to images. Nevertheless, image or spatial distortions still affect the video quality. The inclusion of temporal effects with the spatial effects will make the overall distortion becomes perceptually less or more severe to the observer depending on when the temporal effects appear in a video. A new VQA method, dubbed Modified-MSE (Mod-MSE) is proposed in this paper. Mod-MSE is an improved version of MSE with the inclusion of temporal effect. From the results, the proposed VQA method gives acceptable performance when evaluated with the LIVE video database.