The Algorithm Of Video Enhancement Based On Dark Channel Prior

When it comes to video enhancement technology, most of the time we focuson obtaining the specific image information from the fixed scene, it alsoclearly and accurately needs to extract the characteristic value from oneimage. For example: to get the plate numbers and the colors of illegalvehicles, to extract the physical characteristics of the suspect in thedetails, pavement condition when driving intelligent automobile, and toobserve whether caused clogging rivers and landslides after earthquake etc.However, outdoor computer vision system is seriously affected by the badweather, such as the low light intensity and the other environmental factors.Once our equipment face to these challenges we mentioned above, the imagequality of course will decrease seriously, lead the function of our equipmentgo useless. People’s acquisition of high quality image also can not besatisfied. Therefore, research and study about image restoration and imageenhancement spring up like mushrooms.Based on the previous works and a lot of valuable results, this paperproposed an algorithm of video enhancement using dark channel prior,whichmethod is based on extensive atmospheric degradation model. The main ideaof the algorithm is: firstly, establish contact between the low-lightintensity video and foggy video, if one performs a pixel-wise inversion oflow lighting video or a high dynamic range image, the results look quitesimilar to hazy video. So all need to do is create an automatic impairmentsource detection module as the input step. After that, we used a system inwhich the core enhancement algorithm is based on Dr. He Kai-ming’s darkchannel prior. We optimize the estimation of the transmission picture andglobal atmospheric light from the original. making them more suitable for video enhancement. To estimate the transmission, this paper chose the guidefilter to replace the original soft matting algorithm, we also increase thebox filter to improve the processing speed of the guide filter. In theoptimization with the global atmospheric light estimation, this paper alsoproposes to replace the mean value of the original method of selecting a pointin the input image as the atmospheric light value of the whole picture. Thesimulation result of timing sequence and the actual testing result weregiven.