METRICS PERFORMANCE COMPARISON FOR COLOR IMAGE DATABASE

Abstract

In this paper, two new full-reference metrics for image quality assessment are presented. They are based on the
\nPeak-Signal-to-Noise Ratio (PSNR) and Universal Quality Index (UQI) modified to take into account the Human Visual System (HVS) properties. Many studies confirm that the HVS is more sensitive to low frequency distortions rather than to high frequency ones. It is also very sensitive to contrast changes and noise. To test the effectiveness of the proposed metrics, we have performed a subjective experiment. In our experiment we have used seven types of distortions with three distortion levels in each. Three independent groups of observers from Finland, Ukraine and Italy, have been participating in image quality evaluation test. The analysis of the results shows that the proposed PSNR-HVS provides better correlation to Mean Observer Score (MOS) than PSNR, and that the proposed UQI-HVS metric outperforms two recently proposed state-of-the-art metrics: UQI and MSSIM.