A perceptual PSNR based on the utilization of a linear model of HVS, motion vectors and DFT-3D

In this paper we propose two different methods to obtain a video sequence PSNR taking into account of the subjective quality evaluation of the final users. These proposed "subjective" PSNRs are based on the spatio-temporal human sensitivity method suggested by Z. L. Budrikis [3] and on the experimental results given by Kelly [1] and Robson [2]. The first metric (WPSNR1) consists in a PSNR weighed by the Motion Vectors typical of MPEG 1-2 [4] algorithms; the second one (WPSNR2) is obtained by mean of the Discrete Fourier Transform applied to a three dimensional domain. Experimental results show that these two WPSNRs are more related to the subjective quality perceived by the final user than the usual PSNR.