Image denoising system based on low-rank theory

Abstract

The present invention discloses an image denoising system based on a low-rank theory. An airspace image block matching technology is employed to preliminarily find image similarity blocks, performing matching of the image similarity at an SVD domain, improving the matching precision, and finally confirming the similarity blocks of a target image. According to the low-rand features of the similarity block, a matrix formed by the similarity blocks is subjected to singular value decomposition. About the weight determination problem of weight calculation of the singular value, the image noise intensity is considered, the complex degree of image details is considered, and the noise intensity and an entropy are employed to commonly determine the weight so as to finally realize the better denoising effect. The better matching effect can be obtained, the weight concept is utilized when the solution of a low-rank model, and the setting of the weight depends on the size of the singular value and also depends on an image entropy to allow the setting of the weight to consider the image details and the whole structure, obtain more accurate weight setting and obtain the denoising performance being higher than a general algorithm.

Description

technical field [0001] The invention belongs to the field of image denoising and relates to the low-rank matrix recovery problem based on the low-rank sparse theory. The purpose of denoising is achieved by matching image similar blocks in the space domain and transform domain and solving the low-rank model. Background technique [0002] Vision is the most advanced sensory organ of human beings, so there is no doubt that images play the most important role in human perception. Image processing is to process image information to meet the requirements of human vision and practical application. Noise can be understood as various factors that prevent human visual organs or system sensors from understanding or analyzing the received source information. However, the actual image is often polluted by noise during the process of acquisition, acquisition and transmission, and becomes a noisy image that affects the visual quality. The noise in the image seriously affects the subseque...

AI Technical Summary

Problems solved by technology

However, the actual image is often polluted by noise during the process of acquisition, acquisition and transmission, and becomes a noisy image that affects the visual quality.

Images