Composite degraded image high-quality reconstruction method based on conditional generative adversarial network

Abstract

The invention discloses a composite degraded image high-quality reconstruction method based on a conditional generative adversarial network. The method is based on a conditional generative adversarialnetwork to perform high-quality reconstruction on a composite degraded image in outdoor visual systems such as unmanned aerial vehicle aerial photography, video monitoring and intelligent transportation, and comprises an overall process, establishment of a composite degraded image sample library, establishment and training of a network model, and high-quality reconstruction of the composite degraded image. Unified high-quality reconstruction is carried out on composite degraded images obtained by outdoor vision systems such as unmanned aerial vehicle aerial photography, video monitoring and intelligent transportation through a conditional generative adversarial network. According to the invention, a scheme for establishing a corresponding clear-composite degraded image sample library is provided.; a conditional generative adversarial network is adopted to establish a composite degraded image high-quality reconstruction method, and unified reconstruction of composite degraded images with haze, blurring, a compression effect and the like can be completed. And a light network is adopted, so that the image reconstruction speed is increased, and the application of the method in practice is facilitated.

Description

technical field [0001] The invention belongs to the field of digital image processing, in particular to a high-quality reconstruction method for composite degraded images of an outdoor vision system based on Conditional Generative Adversarial Nets (cGAN, Conditional Generative Adversarial Nets). Background technique [0002] Images collected by outdoor vision systems such as UAV aerial photography, video surveillance, and intelligent transportation are usually affected by various degradation factors such as fog and haze, blurring, and compression effects. These factors are randomly combined in a complex way, resulting in serious degradation of image quality, which not only affects the subjective visual effect of the human eye, but also brings great obstacles to the full use of the outdoor visual system. [0003] The classic atmospheric light scattering model formed by fog and haze images is shown in the formula, where I(x) is a foggy image, J(x) is a clear image, and t(x) is...

AI Technical Summary

Problems solved by technology

[0014] However, the original GAN ​​network does not need to be pre-modeled in the unconditional generative model, which will cause the generator to be uncontrollable. For high

Images