De-noising method based on recessive low-rank structure inside and among nonlocal similar image blocks

Abstract

The invention relates to a de-noising method based on recessive low-rank structure inside and among nonlocal similar image blocks. The method mainly comprises steps of: dividing a target image into subblocks with overlapped structure, decomposing each image subblock into a low-rank matrix and a sparse matrix by means of affine transformation; searching similar low-rank matrixes, transforming each low-rank matrix into a vector and enabling the vectors to form a new data matrix, and obtaining the low-rank structure among the image subblocks by using a rapid singular value truncation method; and performing affine transformation on the obtained low-rank data matrix to obtain a result indicative of the de-noised original image subblocks, averaging the overlapped areas of different image subblocks so as to obtain a de-noising result of the overall image. The invention provides the effective and universal image de-noising method. An experimental result shows that the de-noising method is more effective and robust than other classic image de-noising algorithms and has good application prospect.

Description

technical field [0001] The invention relates to the fields of image processing and computer vision, in particular to a denoising method based on the implicit low-rank structure inside and between blocks of non-locally similar image blocks. Background technique [0002] Image denoising is an extremely important and widely researched hotspot in the field of image processing. The purpose is to remove various noise pollution in the image while maintaining the structural features of the image such as edges and textures. The effect of image denoising directly affects the subsequent image processing work, and eliminating image noise is of great significance to the research of image processing. [0003] In general, image denoising algorithms can be divided into two categories: local methods and non-local methods. The local method is to use some kind of kernel to perform convolution operation with the image, and use all the pixels in the neighborhood where the current pixel is locat...

AI Technical Summary

Problems solved by technology

Although the above-mentioned non-local denoising method based on the similarity of image blocks has achieved good denoising effect, there are also obvious deficiencies, mainly in: (1) the internal structure of the image block has not been effectively utilized; (2) Can only handle Gaussian noise

Images