Research On Image Defogging Algorithm Based On Stereo Vision

Images acquired in natural scenes are affected by the environment and are prone to various noises,the most common of which is image blur caused by haze.The main image dehazing algorithm has a single image defogging method and multiple image defogging algorithms.The most commonly used method is the dark primary color dehazing algorithm.The algorithm is a single image dehazing algorithm.It has a fast speed,high accuracy and is easy to apply.However,it is not good for defogging in bright areas.Color shift.With the popularity of binocular cameras,there are more and more stereoscopic images,and there is almost no defogging algorithm for stereoscopic images.In fact,the multi-image algorithm can greatly improve the accuracy of image depth information estimation because it can obtain the information of the same region from different angles,thus obtaining more reliable smog segment information,which is a great help for accurate defogging.This article is an attempt to use the binocular image for image defogging.In the first part,this paper attempts to improve the dark primary color algorithm,and analyzes the correlation between each color channel and depth image by sub-channel,and obtains a more appropriate depth map.The fog image is then processed using a modified depth map,and the haze of different concentrations is separately defogged to obtain the final result.The improved blue channel dehazing algorithm does not require a smooth depth map to control the color shift of the sky region.It is faster and more accurate than the dark primary color defogging algorithm.However,even if the method is superior to the original method in defogging effect,it produces a slight color shift when dealing with some unnatural,single-tone,over-bright objects.The construction of the depth map of the algorithm is far from the actual depth.The most efficient solution is to use multiple images to try to calculate the exact depth of field of the image in the image domain.For example,the stereo matching algorithm calculates the depth of field of an object by using differences in boundaries,colors,and the like of the same object in two or more images.In the second part,we are inspired by the stereo matching algorithm and the dark channel dehazing algorithm.Try to combine the advantages of both,and propose an image defogging algorithm based on stereo vision.The algorithm uses the Markov random field model to obtain the depth map with more accurate depth information based on the image brightness and gradient information.Then the blue channel algorithm is used to modify the boundary of the depth map to obtain the corrected depth map.Taking the figure as auxiliary information,the atmospheric scattering physical model is used as the algorithm basis to estimate the final fog-free image.Finally,the effectiveness of the model is tested by testing a large number of haze images.