Image denoising method and device, electronic device, and storage medium

Abstract

The invention provides an image denoising method. The method includes the following steps: obtaining a target image; segmenting the target image into multiple image blocks; extracting a noise sub-image of each image block among the multiple image blocks, and recording position data of the noise sub-image of each image block; conducting image stitching and synthesis on the noise sub-images of the image blocks, and obtaining a to-be-denoised image of the target image; denoising the to-be-denoised image and obtaining an image that has been denoised; and on the basis of the position data of the noise sub-image of each image block, conducting image synthesis on the denoised image and the target image, and obtaining a denoised target image. The invention also provides an image denoising device and an electronic device. According to the invention, a class distributed operation is realized, only noise image blocks in the image are processed, and the denoising processing speed is greatly increased.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to an image denoising method, device, electronic equipment and storage medium. Background technique [0002] Noise in images seriously affects image processing, such as image segmentation, coding, feature extraction and object detection, etc. For image denoising, the existing algorithms mainly have the following problems: mean filtering, the main disadvantage is to reduce noise while blurring the image, especially at the edges and details, the larger the field, the more severe the blur; median filtering can overcome The image details brought by the linear filter are blurred, but the median filter method is not suitable for some images with many details; the ideal low-pass filter is prone to serious blurring and ringing in the processing process; the filtering based on spatial domain The processors operate on the gray value of the image pixels. Although they can ach...

AI Technical Summary

Problems solved by technology

For image denoising, the existing algorithms mainly have the following problems: mean filtering, the main disadvantage is to reduce noise while blurring the image, especially at the edges and details, the larger the field, the more severe the blur; median filtering can overcome The image details brought by the linear filter are blurred, but the median filter method is not suitable for some images with many details; the ideal low-pass filter is prone to serious blurring and ringing in the processing process; the filtering based on spatial domain The processors operate on the gray value of the image pixels. Although they can achieve the effect of smoothing the image, they will also blur the outline of the image while smoothing the image; the processing method based on the frequency domain mainly uses filters to convert the useful The signal and the interference signal are separated, but in actual situations, the spectrum of the useful signal and the interference signal often overlap, and it is difficult to balance the noise smoothing effect and the image contour

[0003] Bayesian soft threshold denoising method, which divides the image into low-pass coefficients and high-pass, and mainly denoises the high-pass part. Compared with mean filtering and median filtering, it has better denoising effect, but The efficiency of denoising on high-definition pictures will be very slow

[0004] The existing technology cannot quickly achieve denoising for high-definition images

There are two main reasons: First, image processing is a multi-level processing structure. Simply using the Bayesian soft threshold denoising method requires wavelet decomposition of the image first, and then threshold denoising after the decomposition. The whole high-definition image processing is relatively slow

However, the use of mean filtering or median filtering can easily make the image blurred in the details, so that the desired denoising effect cannot be achieved.

Images