关键词: 图像去雾,暗通道先验,Halo效应,Sigmoid函数,容差机制

Abstract: Dark channel prior theory is usually employed to reduce the effect of fog,however,a bad halo effect often exi-sts.After extensive analysis of halo effect,we proposed an effective algorithm to inhibit the halo effect based on Sigmoid function.Firstly,after many observations of images with halo effect,we found the characteristic of halo effect’s location and sturcture,so we constructed a Sigmoid model with directivity.Secondly,we performed the edge detection of transmission image obtained by dark channel prior,and then obtained the coordinates and edge directions of pixels located at the place of sudden change of depth of field.Thirdly,we used the constructed model to determine the precise locations of those pixels,filled halo region with pixel values of non-halo region,and then obtained optimized transmission image.Finally,we could obtain the clear image with weak halo effect by using the optimized transmission image and introducing a kind of tolerance mechanism.Our algorithm’s characteristic consists in only processing the halo region using the constructed model rather than processing the whole image region,so our method avoids such problem as color distortion in no halo region.The experimental results demonstrate that our method is simple and easy,with faster computational speed and better restoration effect for fog image.

Key words: Haze removal,Dark channel prior,Halo effect,Sigmoid function,Tolerance mechanism