Neural network training and image processing methods and apparatuses, electronic device and storage medium

Abstract

Embodiments of the invention provide neural network training and image processing methods and apparatuses, an electronic device and a storage medium. The neural network training method comprises the steps of obtaining a noisy sample image and a corresponding noiseless sample image; based on a neural network, generating a noiseless estimation image corresponding to the noisy sample image; and according to the noiseless estimation image and the noiseless sample image, training the neural network, wherein the neural network is a bilinear neural network. By adopting the technical scheme, the problem of high possibility of image noise over-fitting in neural network training can be avoided, so that the influence of the trained neural network on information in the images is ensured; and the neural network trained by the neural network training method can realize defogging and denoising processing performed for foggy and noisy natural images, and effectively improves the defogging and denoising effect.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of artificial intelligence, and in particular to a neural network training and image processing method, device, electronic equipment, and storage medium. Background technique [0002] With the development of computer technology and image processing technology, image recognition is widely used in many fields, such as video surveillance, face recognition and so on. Image recognition can identify various target objects by processing, analyzing and understanding images. When performing image recognition, the higher the clarity of the image, the higher the accuracy of recognition. [0003] However, the captured image will be affected by the environment and the air, especially in the case of fog, haze, rain and other bad weather conditions, it is impossible to capture a clear image that is easy for subsequent identification. For this reason, the dehazing technology of natural images cam...

AI Technical Summary

Problems solved by technology

This method does not consider the cause of image degradation, and has a wide range of applications. It can effectively improve the contrast of foggy images and improve the visual effect of images, but it will cause a certain loss of information on prominent parts.

Dehazing algorithms based on image restoration, such as dehazing algorithms based on prior information, include: dark channel dehazing algorithm, algorithm that assumes that object shadows and reflectance are locally uncorrelated, etc., these algorithms can get better dehazing effects , but need to use prior information to estimate, and enhance the noise in the original image and other unnatural information in the process of defogging

[0005] It can be seen that the existing natural image defogging methods will have a considerable impact on the information in the original image, and cannot achieve the natural image defogging process well.

Images