MWPSNR: A new image fidelity metric

Researchers in the field of image and video processing use MSE (Mean Square Error) based fidelity metrics to validate their research results. The most popular MSE-based metrics are PSNR (Peak Signal to Noise Ratio) and WPSNR (Weighted Peak Signal to Noise Ratio). When large quantities of data are to be assessed, subjective metric such as MOS (Mean Opinion Score) is not pragmatic since it needs experts and inordinate amount of time. PSNR and WPSNR are independent of Human Visual System (HVS) parameters and hence they are inappropriate scales to measure potential research results. This paper brings out their inappropriateness and propose a new image fidelity metric called Modified Weighted Peak Signal to Noise Ratio (MWPSNR). This metric has been experimentally proven to be better than PSNR and WPSNR.