Improved performance of PSNR metric using SURF saliency map and CSF filter for standard coding H264/AVC: Improved performance of PSNR metric using SURF and CSF

The utilization of human visual system characteristics in measuring the quality of images and video sequences becomes a very important axis of research. The best objective criterion is as close as possible to the judgment of the human being. Our idea is to establish a pretreatment to existing measure with complete reference (PSNR “Peak Signal-to-Noise Ratio”) for the H.264/MPEG-4 (Motion Picture Expert Group) AVC (Advanced Video Coding) standard. We perform a pretreatment using SURF (Speeded Up Robust Features) saliency map and a CSF filter with distance of 500 mm to improve the performance of these metrics. For the evaluation of our proposed model, the database Video subjective “Live” is used to measure the correlation. Performance indicators, namely Pearson (PLCC), the Spearman coefficient (SROCC) and mean squared prediction error (RMSE) indicate that our idea gives better results for the distortion caused by H264 video encoding.