Research On The Algorithms Of Video Image Defogging And Its Design And Implementation In The Video Processing System

In recent years,the use of video surveillance systems has become more and more widespread,and people's requirements for them have become higher and higher.However,the presence of fog often leads to the reduction of the sharpness and the blurring of the picture captured by the video surveillance system.And it also leads to the fact that the overall color tends to grayish white.It cannot reflect the actual situation on the spot,and it largely constrained and influenced the application of image in video surveillance.Therefore,the defogging processing of foggy images and videos has become particularly important.Firstly,this paper studies and compares the classic image defogging algorithms such as Histogram Equalization Algorithm,Homomorphic Filtering Algorithm,Retinex Algorithm,which based on image enhancement,and the Dark Channel Prior Algorithm based on image restoration.And we summarize the advantages and disadvantages of these algorithms.Secondly,this paper improves the Dark Channel Prior algorithm which has more natural visual effects in fog removal.For the optimization of transmission map in the Dark Channel Prior algorithm,soft matting has very high complexity in time and space and guided filter has the issue that fog removal is not completely.So in this paper we use Butterworth Low-pass Filter to optimize the transmission map for good fog removal results and improvement in the speed of the algorithm.For the images processed by the Dark Channel Prior algorithm has dark brightness,this paper adopts nonlinear superposition to achieve image brightness compensation.For the Dark Channel Prior algorithm using minimum filtering to obtain the dark channel and transmission map usually leads to the "halo" effect,this article apply the morphology of corrosion to weaken the "halo" effect and make fog image's edge more natural after defogging.Then,through MATLAB simulation,we compare the algorithms improvement from the aspects of information entropy,average gradient,edge intensity,variance and algorithm time-consuming.The data prove that the improved algorithm in this paper is better compared with the original algorithms.Finally,this paper designs a video processing system which can be applied to the intelligent fire control system on the VS2010 platform.Combining with Aforge.Net framework and Open CV,we design visual interface to realize system function such as video recording,video fogging and video playback.