Non-local image denoising method based on similar block matrix rank minimization

Abstract

The invention discloses a non-local image denoising method based on similar block matrix rank minimization. The method comprises the following implementation steps of: (1) inputting a noisy image with N rows and M columns; (2) estimating noise standard deviation of the noisy image, and setting a parameter according to the standard deviation; (3) calculating Euclidean distances among image blocks by using the DCT (Discrete Cosine Transform) features of the image blocks in a block-by-block manner; (4) sequencing the Euclidean distances in an ascending manner, and selecting front k corresponding samples to form a similarity matrix; (5) carrying out rank minimization approximating on the similarity matrix, thereby obtaining a low-rank matrix; (6) gathering denoised image block sample sets, thereby obtaining a denoised image; and (7) judging whether the iteration number of times is achieved or not, going to the steps (2) to (6) if the iteration number of times is not achieved, and outputting a result if the iteration number of times is achieved. The non-local image denoising method based on similar block matrix rank minimization has the advantage that the edge texture structure information on the reconstructed result is excellently maintained, thereby being applicable to the digital image preprocessing in the fields of medical images, astronomic images, video multimedia and the like.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, in particular to a denoising method for approaching a noiseless matrix with a low rank of a similar block matrix containing noise, which can be used for digital image preprocessing in the fields of medical images, astronomical images, video multimedia and the like. Background technique [0002] Images have become the most commonly used information carriers in human activities. They contain a large amount of information about objects and become the main way for people to obtain original information from the outside world. However, in the process of shooting, sampling, transmission and storage, the image is often disturbed and affected by external noise, which degrades the image and cannot truly reflect the scene, and the quality of the image preprocessing algorithm is directly related to the subsequent image. The effect of processing, such as image segmentation, feature extraction...

AI Technical Summary

Problems solved by technology

However, there are still some shortcomings in this method: the denoising method in the first step of the method uses a hard threshold, which can easily cause the loss of image edges and texture details; the denoising result of the first step in this method is The initial estimate of the second step, if the initial estimate is not accurate, it directly affects the denoising result of the second step

The disadvantage of this method is that the image blocks are denoised by using the mean and variance of each category. Since some similar blocks are not similar enough, the estimation of the mean and variance is not accurate enough, resulting in an unsatisfactory denoising effect.

Images