ABSTRACT
Although the novel Feature-similarity Metric-based Image quality assessment (FSIM) is simple and has been proved to be better than many traditional methods such as the structure similarity metric (SSIM), the Peak Signal-to-Noise Ratio (PSNR), there are still some difficulties in these distorted pictures with the same type and different levels. We cannot distinguish the differences from the objective values among these images perceived through subjective visual. This paper presents a method based on a new statistics we call it Sudoku for feature-similarity metric (Sudoku-FSIM), which keeps the original method advantages, combines with the Human Visual System (HVS), and considers the influence of the whole image on human eyes, finally gives the objective score comprehensively. Experimental results show that the advanced algorithms are more consistent with human visual system, namely in accordance with Difference Mean Opinion Score (DMOS), and can assess the image quality more precisely than the FSIM.
