Image defogging method and system based on style migration network

Abstract

The invention discloses an image defogging method and system based on a style migration network, and belongs to the technical field of digital image processing. The image defogging method based on a style migration network comprises: the convolutional neural network is used for convolution and pooling; extracting content characteristics from the haze image and the generated image; calculating a content loss function of the haze image and the generated image; extracting style characteristics of the haze-free image, calculating a style loss function of the haze-free image and the generated image, merging the content loss function and the style loss function to obtain a total loss function, adjusting network parameters by iteratively optimizing the loss function, and recalculating the generated image by using the adjusted network parameters. According to the image defogging method based on the style transfer network, a large amount of priori knowledge is not needed, human intervention isnot needed, only the image needs to be processed, the application range is wide, and good application and popularization value is achieved.

Description

technical field [0001] The invention relates to the technical field of digital image processing, and specifically provides an image defogging method and system based on a style transfer network. Background technique [0002] Smog is a common atmospheric phenomenon that mainly occurs in autumn and winter. They are disastrous weather phenomena caused by atmospheric particulate matter such as high concentrations of dust, aerosols, and tiny water droplets scattered in the air. On the formation mechanism, fog and haze are not quite the same. Fog is a kind of weather environment formed in the atmosphere close to the ground by the existence of a large number of tiny water droplets and ice crystals in the air. The radius of the atmospheric particles is generally 1 to 10 microns, and the size of the larger particles is much larger than the wavelength of visible light. Therefore, Its scattering effect on visible light of different wavelengths is basically the same, causing the fog to...

AI Technical Summary

Problems solved by technology

Both of them have brought great adverse effects on human production and life. Due to the spread of the light affected by them, the images acquired by the current general digital image acquisition facilities have low visibility and low contrast. Computer vision tasks such as recognition, object tracking, behavior detection, autonomous driving, etc. cause great inconvenience

Generally, the model of this type of method is affected by conditions such as atmospheric conditions and distance depth of field when the image is collected, and the hardware requirements are too high for the defogging of ordinary equipment, and the application range is limited.

Images