PROJECT TITLE :

A largest matching area approach to image denoising - 2016

ABSTRACT:

Given the success of patch-based approaches to image denoising, this paper addresses the unwell-posed problem of patch size selection. Massive patch sizes improve noise robustness in the presence of excellent matches, however can also lead to artefacts in textured regions due to the rare patch impact; smaller patch sizes reconstruct details a lot of accurately but risk over-fitting to the noise in uniform regions. We have a tendency to propose to jointly optimize each matching patch's identity and size for grayscale image denoising, and present several implementations. The new approach effectively selects the most important matching areas, subject to the constraints of the out there knowledge and noise level, to enhance noise robustness. Experiments on customary take a look at pictures demonstrate our approach's ability to enhance on fixed-size reconstruction, particularly at high noise levels, on smoother image regions.