Non-parameterized image restoration method based on partial differential equation and BM3D (Block-matching and 3D filtering)

Abstract

The invention provides a non-parameterized image restoration method based on a partial differential equation and a BM3D (Block-matching and 3D filtering) and belongs to the technical field of image processing. In a conventional BM3D improved method, a parameter estimation is obtained by use of an existing noise parameter estimation method, and is then introduced into an original BM3D method, however, the result is unstable. In the image restoration method, a denoising process and a noise variance estimation process of the conventional BM3D improved are improved by use of the partial differential equation, the method is characterized by closely fusing with pretreatment of the partial differential equation, and the improved BM3D method is enabled to have high precision and stability. The non-parameterized image restoration method based on the partial differential equation and the BM3D can be applied to various fields of image processing.

Description

technical field [0001] The invention relates to a non-parameterized image restoration method based on partial differential equations and BM3D, and belongs to the technical field of image processing. Background technique [0002] The BM3D (Block Matching and 3D Filtering, image block matching algorithm) method is an excellent algorithm in the field of image denoising, and its results have the advantages of high PSNR value, good visual effect, and complete image detail information. However, it also has obvious shortcomings, mainly reflected in the fact that the processing effect of the BM3D method is extremely sensitive to the noise level parameters in the noise image. Specifically reflected in the following aspects: [0003] 1. When the input noise standard deviation is around 40, the PSNR (peak signal-to-noise ratio) value of the BM3D processing result will change abruptly; [0004] 2. When the input noise standard deviation is non-integer, the PSNR value of the BM3D proce...

AI Technical Summary

Problems solved by technology

[0003] 1. When the input noise standard deviation is around 40, the PSNR (peak signal-to-noise ratio) value of the BM3D processing result will change abruptly;

[0004] 2. When the input noise standard deviation is non-integer, the PSNR value of the BM3D processing result will fluctuate violently;

[0005] 3. Accurate noise standard deviation is not the best input parameter for BM3D method

However, due to the above 1 defect, if the true noise standard deviation is around 40, then the input estimated noise parameters may seriously reduce the restoration quality

Due to the above two defects, the result of noise parameter estimation is generally non-integer, and directly bringing it into the BM3D algorithm will also affect the image restoration result

Due to the above 3 defects, even if some high-precision noise level estimation methods are used to obtain accurate noise level values, directly bringing them into the BM3D method may not necessarily result in optimal processing results

Images