Free Energy Adjusted Peak Signal to Noise Ratio (FEA-PSNR) for Image Quality Assessment

详细信息    查看全文

• 作者：Ning Liu ; Guangtao Zhai
• 关键词：Image quality assessment ; Peak signal to noise ration ; Free energy principle ; Linear mapping
• 刊名：Sensing and Imaging
• 出版年：2017
• 出版时间：December 2017
• 年：2017
• 卷：18
• 期：1
• 全文大小：
• 刊物类别：Engineering
• 刊物主题：Electrical Engineering; Microwaves, RF and Optical Engineering; Imaging / Radiology;
• 出版者：Springer US
• ISSN：1557-2072
• 卷排序：18

文摘

Peak signal to noise ratio (PSNR), the de facto universal image quality metric has been widely criticized as having poor correlation with human subjective quality ratings. In this paper, it will be illustrated that the low performance of PSNR as an image quality metric is partially due to its inability of differentiating image contents. And it is revealed that the deviation between subjective score and PSNR for each type of distortions can be systematically captured by perceptual complexity of the target image. The free energy modelling technique is then introduced to simulate the human cognitive process and measure perceptual complexity of an image. Then it is shown that performance of PSNR can be effectively improved using a linear score mapping process considering image free energy and distortion type. The proposed free energy adjusted peak signal to noise ratio (FEA-PSNR) does not change computational steps the of ordinary PSNR and therefore it inherits the merits of being simple, derivable and physically meaningful. So FEA-PSNR can be easily integrated into existing PSNR based image processing systems to achieve more visually plausible results. And the proposed analysis approach can be extended to other types of image quality metrics for enhanced performance.