Single image dehazing based on linear transformation

Abstract

In order to solve the problems of sky distortion, edge artifact and overall dark of the restored image, a single image dehazing algorithm based on linear transformation was proposed. Firstly, adaptive compression normalize the hazy image, the ratio of the minimum channel to the maximum channel of the image is defined as the anti-saturation of the image, and the model of medium transmission and anti-saturation of the image without haze is established. Then, combining the linear transformation model between haze-free image anti-saturation and hazy image anti-saturation, the initial medium transmission is calculated by linear transformation, the medium transmission obtained by different linear transformation rates are weighted fused through the maximum channel map of image with haze, and the final medium transmission is obtained by fast guided filtering. Finally, atmospheric light values are derived from the quad-tree sub-block search method, and then the restored image is obtained. Both subjective and objective results of experimental simulation show that the algorithm has obvious recovery of details, good effect on the hazy images of different scenes, and good restoration effect on sky region.