Blind Quality Metric using a Perceptual Importance Map for JPEG-20000 Compressed Images

JPEG-2000 coding introduces at high compression ratios, some perceptual impairments (blurring and ringing effects), which can be exploited by a no reference (NR) quality metric. Even if these distortions are present in the whole processed image, the human visual system judges, the perceptual quality, identifying and selecting some regions of interest. In this paper, to judge of the perceptual quality of JPEG-2000 compressed images, we propose to measure both of ringing and blurring distortions, locally weighted by an importance map generated on the one hand, by an Osbergers modified model and on the other hand, by a simple algorithm, of attention model. The respective predicted scores have been compared with the subjective quality scores. With a comparative study on the contribution of each importance map, we demonstrate the significant interest of these weights in a NR quality metric.