Improved performance of quality metrics using saliency map and CSF filter for standard coding H264/AVC

Despite the great evolution in the field of quality metrics, classical tools such as the PSNR remain useful in the field of evolution of video coding research. Our idea is to improve the limited performance of these tools using simple and well known techniques. In the quality measurement of images and video sequences, the use of the characteristics of human visual systems becomes a priority in order to have a better correlation between the objective measurement and the judgment of the observers. For this, we try to improve the performance of the two existing metrics (PSNR “Peak Signal / Noise Ratio”, DVQ “Digital Video Quality”) for the H.264 / MPEG-4 (Motion Picture Expert Group) AVC (Advanced Video Coding). This improvement is carried out by extracting the zones of interest using a saliency map and using a filtering by the contrast sensitivity function CSF. We did a study using two types of saliency map SURF (Speeded Up Robust Features) and HARRIS saliency map. We use the subjective video database “LIVE” to test the performance of our proposed idea. Performance indicators, namely Pearson (PLCC), the Spearman coefficient (SROCC) and mean squared prediction error (RMSE) indicate that SURF give best results for the distortion produced by H264 video encoding then HARRIS. Our idea is proven especially in short distances of observation.