Expected Patch Log Likelihood Based on Multi-layer Prior Information Learning
How to preserve the edge and texture details has been a difficult problem in image denoising. In this paper, we propose a multi-layer prior information learning method, which combines the statistical and geometric features of the image to describe the attributes of the prior information more accurately and completely. The experimental results show that our proposed method is superior to the EPLL (Expected patch log likelihood) method with a single statistical characteristic for a priori learning in both visual and quantitative evaluation.
KeywordsImage denoising Image prior Expected patch log likelihood Statistical characteristic Geometric characteristic
- 3.Zoran D, Weiss Y.: From learning models of natural image patches to whole image restoration. In: 2011 IEEE International Conference on Computer Vision (ICCV), pp. 479–486. IEEE (2011)Google Scholar