Abstract

Spatially adaptive nonlocal patch-wise estimation is one of the most promising recent directions in image processing. Within this framework, a number of Block Matching 3-D filtering (BM3D) algorithms have been developed for different imaging problems [1], [2], [3]. In this paper we present the analysis/synthesis frames for BM3D image modeling and use them to develop novel recursive deblurring algorithm based on the augmented Lagrangian technique. In simulation experiments we demonstrate that proposed algorithm essentially outperforms current stateof-the-art methods. 1.

Citations