A Method for Removing Poisson Noise from Images Based on Nonlocal Similarity Low Rank Matrix

Abstract

The invention discloses an image removal Poisson noise based on a non-local similarity low-rank matrix, comprising the following steps: (1) Noise analysis According to the joint probability density function of the Poisson distribution and the maximum likelihood principle, the Poisson noise removal method is obtained. The noise is equivalent to minimizing the KL-divergence function, and then using the prior knowledge of the non-locally similar blocks of the image as a regular term to perform a stable numerical solution; (2) Establishing a model: According to the noise analysis, combined with the posterior probability formula, it is obtained The Poisson low-rank denoising model is as follows: (3) Denoising processing: According to the Poisson low-rank denoising model, through optimization knowledge, the solution F j and the iterative formula of f, and then the final solution is obtained through the alternate iteration method; (4) output the denoising image. The invention establishes a low-rank matrix denoising model according to the non-local similarity of the image, and uses the rank minimization method to not only remove the image noise efficiently, but also retain the detailed information such as the structure, texture and edge of the image as much as possible to obtain better visual Effect.

Description

technical field [0001] The invention belongs to the field of image denoising, in particular to a method for removing Poisson noise from an image based on a non-local similarity low-rank matrix. Background technique [0002] Photon noise is caused by the statistical nature of light and the photoelectric conversion process in image sensors, and is commonly found in images generated by magnetic resonance imaging systems, electron microscopy imaging systems, and astronomical imaging systems. From the statistical point of view, the signal is closely related to the noise, the higher the signal strength, the greater the noise. Since the photon noise obeys the Poisson distribution, the problem of eliminating photon noise can be transformed into the problem of noise removal obeying the Poisson distribution, which will provide convenience for medical, biological, astronomical and other research fields. [0003] Image denoising is a basic topic of image processing and computer vision,...

AI Technical Summary

Problems solved by technology

However, the above methods use different emphases to build models and construct algorithms. While removing noise, they lose details such as image structure, texture, and edges to varying degrees.

Images