Denoising efficiency analysis based on no-reference image quality assessment
Andrii Rubel, Владимир Васильевич Лукин · 2018 14th International Conference on Advanced Trends in Radioelecrtronics, Telecommunications and Computer Engineering (TCSET) · 2018
Visual quality is of great importance for image denoising techniques. Quality assessment, especially by humans, allows knowing whether the use denoising is expedient or not. This paper is focused on analysis of denoising efficiency and expedience using no-reference image quality assessment (NR IQA) and possible ways to predict quality evaluation by humans based on several input parameters. Two denoising techniques, namely the DCT filter and the BM3D filter, are considered. It is shown that Spearman rank order correlation coefficient (SROCC) between values of state of the art objective no-reference image quality metrics and subjective (human) opinions does not exceed 0.7. It is also demonstrated that prediction of image visual quality evaluation by humans for denoising is still challenging.