Image non-local mean denoising method and system

Abstract

The invention provides a non-local mean denoising method for an image, comprising: acquiring an image to be denoised, extracting an image block of a preset size, reading the spatial data of three RGB color components of the image block to form a vector matrix, Component analysis and dimensionality reduction to generate a dimensionality-reduced residual vector matrix; for the image block, according to the preset S-level search area, in the search area, according to the preset search step size of each level of search area, to the image block. The center pixel point is the center from top to bottom, and the preset size of the search window is moved from left to right to obtain the search block; among them, the search step size of the search areas at all levels increases in turn; calculate the residual vector matrix corresponding to the search block and the pixel The Euclidean distance of the vector corresponding to the point, then calculate the weight value of the image block according to the Euclidean distance, and finally obtain the filter value of the center pixel of the image block according to the weight value to complete the image denoising process. The above method has the characteristics of low time complexity and fast denoising speed.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image non-local mean denoising method and an image non-local mean denoising system. Background technique [0002] During the imaging process of digital images and videos, various noises are often introduced, including atmospheric noise, sensor noise of camera equipment, especially for imaging images with low illumination, the noise introduced is often serious. Image denoising techniques include spatial denoising and frequency domain denoising techniques. Frequency domain denoising needs to go through the transformation from space domain to frequency domain, which is often complex, and the traditional Wiener filter denoising effect is not good. Traditional spatial filtering denoising methods, such as mean filtering, median filtering, and statistical sorting filters, often lead to blurred images and poor denoising effects. At present, the spatial filtering method with...

AI Technical Summary

Problems solved by technology

The algorithm complexity of this method is simple, and it can better remove image noise while retaining image details, but its time complexity is high, and the complexity is O(M×N×(2r+1) 2 × L 2 ), where M and N are the length and width of the image respectively, r is the search radius of the current block of the image, and L is the length and width of the image block (the width and length are the same)

As the search radius r increases, the image denoising effect is better, but the time complexity increases rapidly

In practical applications, fast image / video denoising technology is often required to achieve real-time processing, and the high time complexity limits the application of NLM algorithms

[0003] When the image noise is serious, the image details are covered by the noise, which will affect the similarity of the image blocks calculated in the traditional NLM algorithm, and it is impossible to obtain accurate similarity blocks.

At the same time, the research shows that in most cases, the similarity of image blocks will decrease with the increase of the search radius, the weight of the search blocks at a longer distance is lower, and the search step time complexity of the traditional NLM algorithm is higher

Images