Adaptive patch selection method for a single image deblurring
Yujin Yun, Jonghyun Bae, Jae‐Seok Kim · 2011
This paper proposes an adaptive patch selection method for a deblurring algorithm based on the variational Bayesian theory. In deblurring process, only small portion of an image is used due to the complexity issue. By searching the entire image, a patch with the most edge information is selected. Depending on the feature of each image, an appropriate shape of a patch is selected. Saturation regions and large uniform regions are avoided since they degrades blur kernel estimation process. Compared to the other methods, proposed method chooses the most proper part of an image and restores high quality images.