Single image defogging method based on generation confrontation network

Abstract

The invention provides a single image defogging method based on a generation confrontation network. The method includes the steps of 1, constructing a generator network model, inputting a training sample set subjected to fogging processing into the generator network model, and generating preliminary defogged images imitating fog-free images in a test sample set; 2, constructing a decider network model, inputting the preliminary defogged images into the decider network model, calculating a cost function, and performing the following steps that if a calculation result of the cost function is smaller than a preset defogging threshold value, the input images are judged as the fog-free images of the test sample set, and the generator network model is used as an optimal training model; otherwise, the input images are judged as the preliminary defogged images generated by the generator network model, tensorflow is used for training the generation confrontation network, and step 2 is repeatedly executed; 3, inputting a foggy image set into the optimal training model and outputting images after defogging. The method has the advantages of scientific design, high practicability, simple operation and a great defogging effect.

Description

technical field [0001] The invention relates to the technical field of single image defogging, in particular to a method for defogging a single image based on a generative confrontation network. Background technique [0002] In haze weather, there are many atmospheric particles in the air. These particles not only absorb and scatter the reflected light of the scene, but also scatter some atmospheric light to the camera, causing the image acquired by the camera to deteriorate, making the image low in contrast, poor in visibility, and severely degraded in quality. [0003] At present, image defogging algorithms can be mainly divided into three categories: the first category is based on image enhancement, but based on image enhancement, some information features of the image will be lost. The second category is image restoration based on physical models. The purpose of image restoration algorithms is to obtain natural and clear images with good visibility while maintaining goo...

AI Technical Summary

Problems solved by technology

These particles not only absorb and scatter the reflected light of the scene, but also scatter some atmospheric light to the camera, causing the image acquired by the camera to deteriorate, making the image low in contrast, poor in visibility, and severely degraded in quality

[0003] At present, image defogging algorithms can be mainly divided into three categories: the first category is based on image enhancement, but based on image enhancement, some information features of the image will be lost

The second category is image restoration based on physical models. The purpose of image restoration algorithms is to obtain natural and clear images with good visibility while maintaining good color restoration performance; to establish the physics of atmospheric scattering based on the deterioration of images under blurred conditions. model, first of all, it is necessary to estimate the physical parameter model, such as atmospheric illumination intensity and transmittance (depth), and then invert the physical model to obtain a fog-free image, but the image restoration processing range based on the physical model is limited; the third type is based on deep learning Image dehazing algorithm, such as convolutional neural network applied to image dehazing