Foggy image sharpening method and device and storage medium

Abstract

The invention discloses a foggy image sharpening method and device, and a storage medium. The method comprises the steps of obtaining a foggy image data set and a fogless image data set; constructinga cyclic generative adversarial network and a perception loss network, wherein the cyclic generative adversarial network comprises a first generator, a second generator, a first discriminator and a second discriminator; inputting the foggy image data set and the fogless image data set into a cyclic generative adversarial network, and training in combination with the perception loss network to obtain an optimal generation model; inputting the image to be defogged into the optimal generation model to obtain a corresponding predicted fog-free image; and performing Laplace pyramid reduction on thepredicted fog-free image to obtain a clear fog-free image. According to the method, the characteristics of the foggy image can be automatically extracted, style conversion of a single foggy image anda single fogless image is completed, and the foggy image and the real fogless image which are matched in pairs in the same scene do not need to be obtained for training, so that the foggy image is conveniently and flexibly clarified.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method, device and storage medium for clearing foggy images. Background technique [0002] Clearing a single foggy image is to take a series of methods to remove the interference of fog in the image, and then restore a high-definition image. [0003] Traditional foggy image clearing methods are divided into two categories: methods based on prior conditions and methods based on deep learning. Dehazing of a single image based on prior conditions refers to the use of prior information to estimate the parameters of the atmospheric scattering model. , because the assumptions of these prior conditions do not always hold true in specific scenarios, so prior-based methods cannot achieve foggy image clarity well in some cases. Most CNN foggy image clearing models based on deep learning need to evaluate the intermediate parameters of the atmospheric scattering model, an...

AI Technical Summary

Problems solved by technology

[0003] Traditional foggy image clearing methods are divided into two categories: methods based on prior conditions and methods based on deep learning. Dehazing of a single image based on prior conditions refers to the use of prior information to estimate the parameters of the atmospheric scattering model. , because the assumptions of these prior conditions are not always true in specific scenarios, so the method based on the prior cannot achieve foggy image clarity in some cases.

Most CNN foggy image clearing models based on deep learning need to evaluate the intermediate parameters of the atmospheric scattering model, and need to input pairs of foggy images and corresponding ground truth images, but at the same time obtain pairs of foggy images in the same scene and real fog-free images are also more difficult

Therefore, the traditional foggy image clearing methods have certain limitations.

Images