Double-residual denoising method based on attention distribution mechanism

Abstract

The invention discloses a double-residual denoising method based on an attention distribution mechanism, and the method comprises the steps: constructing a training data set, and carrying out the preprocessing of the training data set; constructing a network denoising model by using an attention distribution mechanism and a convolutional neural network of a double-residual network structure; setting a hyper-parameter and a loss function of the network denoising model, and optimizing the loss function; adding different levels of noise to the training data set and training to obtain a trained network model; image denoising is carried out on the trained network model, the noise image is evaluated through the structural similarity and the peak signal-to-noise ratio index, and the effect of improving the denoising performance and the imaging quality is achieved.

Description

technical field [0001] The invention relates to the technical fields of computer vision and image processing, in particular to a double residual denoising method based on an attention distribution mechanism. Background technique [0002] The image is stored in the computer as a digital, but due to the influence of the environment, equipment, human factors, etc. during the collection and transmission process, noise is inevitably introduced, which affects the readability of the image, resulting in lower image quality. This phenomenon is called degradation. In order to solve this degradation phenomenon, image denoising technology is proposed. Denoising is one of the most important methods to improve image quality. At the same time, denoising is convenient for subsequent image classification, segmentation and recognition tasks. [0003] Image denoising is a classic problem in the field of image processing, and it is also an important step in computer vision preprocessing. Altho...

AI Technical Summary

Problems solved by technology

Although the traditional image denoising algorithm can achieve better denoising effect, there are still two disadvantages: 1. These methods usually involve complex optimization problems, and the calculation time is slow

2. Most of the models are non-convex, and parameters need to be manually set during the optimization and adjustment process, so it is difficult to obtain good image restoration effects efficiently

[0004] In recent years, the image denoising method based on deep learning has been well developed in academic research. At the same time, the attention mechanism in deep learning is also being continuously improved, and the image denoising network combined with the attention mechanism still exists. There is a certain room for improvement, which includes: the existing denoising network is mainly realized by local convolution, and the receptive field is small; for the detail texture part, there is no excessive attention, and the image after denoising is prone to edge smoothing

[0005] The publication number in the prior art is: CN113610719A, and the name is: a kind of attention and densely connected residual block convolution kernel neural network image denoising method, and its disclosed network denoising model includes input layer, hidden layer, convolution Layer and output layer, but firstly, the image augmentation method is relatively simple in the training process, and secondly, the attention mechanism is globally consistent, and the problem of attention distribution is not considered

Images