PERCEPTUAL QUALITY ASSESSMENT OF DENOISED IMAGES
Kai Zeng, Zhou Wang · 2015
Image denoising has been an extensively investigated problem in the field of image processing, but little research has been dedicated to the development and validation of image quality assessment (IQA) approaches for denoised images. Without such IQA methods, fair comparison is difficult and further improvement is aimless. In this study, we first create a denoised image database and conduct a subjective experiment to compare the quality of these images. We find widely used IQA measures only have moderate correlations with subjective opinions. Furthermore, we propose a novel objective IQA approach that combines the full-reference SSIM approach with natural scene statistics (NSS) based reduced-reference IQA methods. Experimental results show that the proposed scheme outperforms state-of-theart IQA models.