An image denoising method based on multi-scale parallel CNN

Abstract

The invention discloses an image denoising method based on multi-scale parallel CNN, which includes five steps. Step 1, build a multi-scale parallel convolutional neural network model, in which there are only convolutional layers and activation layers, and add residual learning at the same time; Step 2, set the training parameters of the multi-scale parallel convolutional neural network model; Step 3, select training set, and perform operations such as cropping and flipping on the selected training images to increase the number of training sets; Step 4, select the mean square error as the loss function, and train the multi-scale parallel convolutional neural network model by minimizing the loss function, Obtain an image denoising model; Step 5, input a noise image of any size into the image denoising model, and the output is a clean image after denoising. The invention can preserve the edge information and detail information of the image as much as possible while denoising, can improve the structural similarity of the image, and obtain a high-quality denoising image.

Description

technical field [0001] The invention relates to the fields of computer vision and digital image processing, in particular to an image denoising method based on multi-scale parallel CNN. Background technique [0002] Due to the inevitable influence of the surrounding environment, equipment, human factors, etc. in the process of image acquisition, the obtained image is always noisy, and the noise will deteriorate the quality of the image, thus affecting the readability of the image and the image subsequent processing. The task of image denoising is to remove the noise from the image to reduce the influence of noise on the image. At present, there are many classic methods for image denoising, but they can be roughly divided into two categories, one is based on spatial domain filtering, such as mean filtering, median filtering, etc.; the other is based on transform domain filtering, such as Gaussian scale mixture model Bayesian least squares in . These local filtering methods...

AI Technical Summary

Problems solved by technology

These local filtering methods neither filter in the global scope nor take into account the connection between natural image blocks and blocks, so the denoising effect obtained is not satisfactory

Images