Elliptical search window and parameter self-adaption non-local mean value denoising method
A non-local mean, search window technology, applied in the field of image processing, can solve the problem of the degradation of the ability to retain the details of the image, the distinction and analysis of pixels that cannot have similar features, etc.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Publication Date
- 2018-11-06
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
[0001] technology neighborhood
[0002] The invention relates to the technical field of image processing, in particular to an image denoising processing method, which can be used to process Gaussian noise existing in natural images. Background technique
[0003] In the direction of noise suppression and denoising of digital images, the non-local mean denoising algorithm (NLM) uses the property of Gaussian white noise with zero mean value and the characteristics of other similar image blocks in the image, by searching for other similar image blocks in the local neighborhood. Similar pixels are weighted and averaged to estimate the true value of the target pixel. This algorithm has a good denoising effect and the ability to preserve image details. Since it was proposed, many improvement methods have been proposed to improve the performance of NLM. The improved factors include: computational efficiency, the shape of the search window, The size of the image block is adaptive, the...
Examples
Embodiment
[0033] An elliptical search window and parameter adaptive non-local mean denoising method provided in this embodiment specifically includes the following steps, see figure 1 :
[0034] (1) Input a 512×512 Monarch noise image I, and the noise level is σ=20.
[0035] (2) Traverse the pixel point i in the noise image I point by point. In this example, the pixel point at the image (357,354) is selected as
[0036] For an example, see figure 2 , get the local neighborhood Ω centered on the pixel i , the neighborhood size is 11×11.
[0037] (3) Calculate the square local neighborhood Ω by the following formula 1 i Gradient image G along horizontal and vertical directions i :
[0038]
[0039] Then, for the gradient image G i Perform singular value decomposition to get the eigenvalue S i =(S i,1 ,S i , 2 ) and the eigenvector V i =(V i,1 ,V i , 2 ),which is
[0040] The structure tensor C is calculated by the following formula 2 i :
[0041]
[0042] in, ...