A new perceptual quality metric for compressed video

This paper presents a new video quality metric for automatically estimating the perceptual quality of compressed video sequences. Distortion measures such as the mean squared error (MSE) and the peak signal to noise ratio (PSNR) have been found to poorly correlate with visual quality at lower bit-rates. The proposed quality metric (MOSp) predicts perceptual quality of compressed video using sequence characteristics and the mean squared error (MSE) between the original and compressed video sequences. The metric has been tested on various video sequences compressed using the H.264 video compression standard at different bit-rates. Results show that the proposed metric has better correlation with subjective quality compared to popular metrics such as PSNR, SSIM and PSNRplus. The new metric is simple to compute and hence suitable for incorporation into real-time applications such as the standard video compression codecs inorder to improve the visual quality of compressed video sequences.