Image denoising method based on residual learning and convolutional neural network

Abstract

The invention discloses an image denoising method based on residual learning and a convolutional neural network. The image denoising method comprises the following steps: firstly, denoising an input noisy image by adopting a convolutional neural network method and a residual learning method respectively; denoising the noisy image by adopting a method of combining the convolutional neural network method and the residual learning method, adding padding into the convolutional neural network, performing batch normalization operation, adding a spanning connection structure from a shallow layer to adeep layer into the network, and performing training by adopting the convolutional neural network of an Adam algorithm; and finally, outputting the denoised image. The network depth can be expanded,network degradation and loss and loss of information in a transmission project are effectively avoided, and the maintaining effects of the depth and structural information of a convolutional neural network denoising model are improved.

Description

technical field [0001] The invention relates to image processing technology, in particular to an image denoising method based on residual learning and convolutional neural network. Background technique [0002] In the process of image acquisition, acquisition, encoding and transmission, some noise will be introduced due to various reasons, resulting in a serious decline in image quality. Therefore, image denoising and image quality improvement are important research parts in image processing technology. Through some noise processing techniques, image noise can be removed as much as possible, so as to obtain information more efficiently, which is conducive to further processing of image feature extraction, signal detection, and image compression. However, some traditional denoising methods always cause loss of information during the transmission process. When using convolutional neural networks for image processing operations, the preprocessing operations required for input i...

AI Technical Summary

Problems solved by technology

[0003] However, the convolutional neural network cannot maintain the image structure information well, and in the process of deepening the network depth, the training difficulty will increase and the training effect will decrease. Therefore, it is urgent to conduct research on related content to improve the convolutional neural network structure. information retention

Images