A Method of Removing Poisson Noise Based on Deep Convolutional Neural Network

Abstract

A method of removing pine noise based on deep convolutional neural networks proposed in the present invention. The main contents include: network architecture, training network, supervision framework, and removing pine noise., Through the deep convolutional neural network DET, each layer uses a 3 × 3 convolution nucleus on the previous layer output, and then extracts the clear image of the original picture and the input image combination to generate the original image, generating the original to the original to the original.The estimation of clear images, the final output income clear image.The present invention breaks its dependence on the data model and is easy to adapt to a certain data type by training. In addition, high -level parallelization can be used for fast operations on the GPU, so that it can get more accurate images and promote the fields of medicine and astronomy and other fieldsResearch and development of subsequent work.

Description

technical field [0001] The invention relates to the field of image denoising, in particular to a method for removing Poisson noise based on a deep convolutional neural network. Background technique [0002] Image denoising is often used in video surveillance, medical, astronomical images and other fields to restore noisy images and retain key image information, that is, to remove factors that affect the understanding and analysis of image source information, and to obtain clearer visual effects. Specifically, in the security field, in the fuzzy video, the outline of the target in the frame can be clear to help identify specific people or objects. In medicine, the complexity of medical imaging systems often produces noise, which leads to a decrease in the quality of medical images, which in turn affects medical analysis and diagnosis. The accuracy of system analysis, so removing Poisson noise can enable medical analysis and diagnosis to obtain accurate images, which is benefi...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that existing methods need more effective comparisons to promote research, the purpose of the present invention is to provide a method for removing Poisson noise based on a deep convolutional neural network. By constructing a deep neural network, a supervised method is adopted to learn to remove Poisson noise. Loose noise, easy to adapt to a certain data type through training, in addition, highly parallelizable can be quickly calculated on the GPU, making it possible to get more accurate images faster, and promote the research and development of follow-up work in the fields of medicine and astronomy

Images