Image defogging method based on adversarial neural network

Abstract

The invention provides an image defogging system based on an adversarial neural network, and the invention comprises the following steps: S1, selecting an image data set of RGBD, and making a defogging data set through employing an atmospheric scattering model; s2, normalizing the size of the picture in the data set to a * a; s3, building an adversarial neural network defogging model, wherein the model is divided into two parts: a generative network and a discrimination network; s4, training the adversarial neural network model by using the data set; and S5, storing the trained model, and inputting a foggy image to obtain a clear image. The invention does not need to manually extract features, effectively avoids intermediate variable prediction errors, realizes end-to-end defogging, and is simple and wide in applicability.

Description

technical field [0001] The present invention mainly relates to the field of image processing, in particular to an image defogging system based on an adversarial neural network. Background technique [0002] In haze weather, there are a lot of suspended particles such as water vapor and dust in the air. They absorb and scatter light, resulting in severe color attenuation, decreased clarity and contrast, and poor visual effects in the pictures collected by the equipment, which have a serious impact on subsequent computer vision tasks. Therefore, effective dehazing of hazy images is necessary. [0003] In recent years, the research on image dehazing algorithm has made great progress. At this stage, image dehazing research is mainly divided into two types, feature-based and prior-based methods and learning-based methods. Dehazing algorithms based on features and priors focus on the estimation of transmission maps, and the difficulty lies in the selection of features and prior...

AI Technical Summary

Problems solved by technology

[0004] In fact, the above methods all have flaws. There will be errors when predicting intermediate variables based on feature and prior methods and step-by-step learning algorithms. When calculating the formula according to the atmospheric scattering model, the error still exists

The end-to-end dehazing method can avoid these defects, but there are problems that the model cannot be explained and the network training is difficult.

Images