Composite regularized image denoising method in combination with nonlocal prior

Abstract

The invention discloses an image multi-feature fusion-based non-local mean denoising method. The method improves a method for calculating similarity between image blocks in a non-local mean denoising method, and belongs to the field of research of image denoising in image processing; and in a conventional non-local mean method, the similarity between the blocks is measured by adopting a Euclidean distance or a Gaussian weighted Euclidean distance, so that certain defects exist in the mode, dissimilar image blocks are easily introduced, errors are accumulatively caused, structure information of an image cannot be well kept especially in a texture detail region, and the denoising performance is reduced. For the problems, the similarity between the image blocks is calculated according to an LBP texture feature, and the hybrid similarity is calculated in combination with the LBP feature and a grayscale feature simultaneously, so that weight allocation of the similar image blocks is finally determined. The image multi-feature fusion-based non-local mean denoising method is remarkably improved in denoising effect.

Description

technical field [0001] The invention relates to the technical field of image denoising, in particular to a composite regularized image denoising method combined with non-local prior. Background technique [0002] In the process of image formation, transmission and recording, it will inevitably be disturbed by noise. The introduction of noise not only reduces the quality of the image, but also seriously affects the subsequent processing of the image. Therefore, image denoising has become a basic and important step in image processing. On the basis of preprocessing, improving image quality can provide a more reliable and realistic basis for subsequent image processing. The ultimate goal of image denoising is to better preserve important structural information such as edges and textures in the image while removing noise. [0003] The basic idea of ​​the regularized image denoising method is: introduce the prior information of the original image into the objective function as ...

AI Technical Summary

Problems solved by technology

[0014] The purpose of the present invention is to provide a compound regularized image denoising method combined with non-local priors, which solves the problem of losing texture and other detailed information in TV model denoising, and the problem of "block effect" appears

Images