Convolutional neural network self compression-based image denoising method and system

Abstract

The invention discloses a convolutional neural network self compression-based image denoising method and system. The image denoising method comprises the steps of inputting a to-be-denoised image to afirst neural network model; performing neural network implicit information extraction processing on the to-be-denoised image by adopting a convolutional layer in the first neural network model to obtain image contour information of the to-be-denoised image; performing dimension reduction sampling processing on the image contour information by adopting a dimension reduction sampling layer in the first neural network model, and finally outputting an underlying output image; inputting the underlying output image to a second neural network model; performing image up-sampling processing on the underlying output image in the second neural network model, performing interpolation processing by adopting most adjacent pixel points simultaneously in the image up-sampling processing process, and outputting a restored denoised image; and taking the finally output restored denoised image as a denoised image. According to the method and the system, the image denoising quality and efficiency can be ensured.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to an image noise reduction method and system based on convolutional neural network self-compression. Background technique [0002] In today's society, due to the increasingly popular research and development of computer vision, the demand for various image processing algorithms such as image recognition and image classification has also risen sharply. However, as the input of computer image processing, the quality of an image will largely affect the later algorithm operation. The quality of an image is caused by the noise points on the image, so how to reduce or even eliminate the noise points on the image has become a very hot topic [0003] However, the existing image denoising or noise reduction technologies are not perfect, and the algorithm complexity, effectiveness, and time consumption are still at a level that urgently needs to be optimized. The huge imag...

AI Technical Summary

Problems solved by technology

[0003] However, the existing image denoising or noise reduction technology is not perfect, and the algorithm complexity, effectiveness, and time consumption are still at a level that urgently needs to be optimized.

The huge demand for images requires more complete and richer technical support to be filled, but the current technical model cannot complete large-scale image noise reduction in a short period of time, let alone scientifically and reasonably reduce the noise points on the image. The removal of quality, so that it cannot meet the function of computer vision post-sequence algorithm processing and recognition

Images