A depth convolution neural network image denoising method based on Inception model

Abstract

The invention relates to the technical field of image processing, more specifically, to a depth convolution neural network image denoising method based on an Inception model. Building a training network with one-layer Inception model; setting the parameters of network solution and choosing the appropriate supervisory framework; the parameters of the network structure are optimized, and the obtained network is used for denoising. The depth convolution neural network image denoising method based on the Inception model of the invention is based on the depth convolution neural network and combinesGaussian white noise to simulate unknown real noise and the network containing the Inception model to reduce the parameter quantity on the basis of maintaining the original denoising effect; throughthe way of data expansion, the content of training data is enriched, and the neural network can learn the inner structure of image distribution or noise. By introducing the residual structure, the network converges more easily and the effect is the best.

Description

technical field [0001] The present invention relates to the technical field of image processing, and more specifically, to an image denoising method based on a deep convolutional neural network based on an Inception model. Background technique [0002] With the deepening of the digital revolution, digital images have become an indispensable part of people's lives. In the process of acquiring images, due to the limitations of the imaging system, the acquired images are often noisy; for example, thermal noise due to the internal resistance of the electronic components of the imaging system will be affected by temperature; remote sensing satellite images, because space Random noise due to electromagnetic interference. When the noise intensity reaches a certain level, the image quality will be severely degraded. This makes subsequent information dissemination and image processing difficult. Therefore, image denoising technology is an indispensable research topic in the field ...

AI Technical Summary

Problems solved by technology

The BM3D in the non-local algorithm makes full use of the self-similarity in the natural image and has a good denoising effect; however, when the image noise intensity increases, the useful information that can be used inside the image decreases, and the denoising is performed according to the information inside the noise. The denoising effect of the noisy BM3D method will become worse

The image denoising based on the deep learning model is different from the traditional image denoising technology. It is based on setting network parameters and performing network training for specific noise types and noise strengths to learn the internal characteristics of image blocks or noise blocks. Noise removal; however, these algorithms have many deficiencies: the denoising algorithm based on the spatial domain has a sharp decline in denoising performance when the image contains low noise; the denoising algorithm based on the frequency domain, when the threshold is unreasonable Domain image denoising often blurs edges and some high-frequency texture information due to the ringing phenomenon, which is lost during frequency-domain transformation mapping, causing the image to lose part of the high-frequency information; image denoising method based on wavelet changes , because of the high complexity of the image itself, it is difficult to select a suitable wavelet base to separate noise and signal, and it requires experience.

Images