Research On Defogging Algorithm And Its Application In Video Monitoring

The frequent occurrence of haze weather makes a great challenge to the computer vision imaging system.The image clarity has a serious decline by the haze,which directly leads to a significant reduction in the identifiable information of images and influences the target judgment and recognition,brings great difficulties to the subsequent processing.In recent years,the defogging algorithm has received a great attention.The dark channel prior defogging algorithm is considered to be one of the most excellent algorithms.But it still can not be widely used because of its high complexity.For this problem,a fast defogging algorithm based on dark channel prior was proposed.The proposed algorithm can effectively reduce the complexity,so that the processing reaches quasi-real-time.Finally,the proposed algorithm was applied to a new video monitoring system based on HTML5,which proved the practicability of the improved algorithm.The main work of this paper is as follows:1.This paper briefly introduced the research status of image defogging around the world,including image enhancement class and physical model class.Then some basic knowledge of image processing was introduced,including how digital images stored in a computer,several common color space models and atmospheric scattering model.Meanwhile,several classic defogging algorithms were also introduced.Their principles were analyzed in detail and verified experimentally.2.The improvement of dark channel prior defogging algorithm.First,the principle of dark channel prior was introduced in detail,and its disadvantage that the algorithm complexity is too high was pointed out.Then we introduced three improved algorithms for improving the transmittance,analyzing their principles and verifying effects of their refinements.The whole defogging process has two key steps with high complexity.One is computing the dark channel value.The other is refining the transmittance.So we proposed two improvements for both steps.The speed of computing the dark channel value was accelerated by blocking the image.At the same time,the block effect was eliminated through the linear interpolation algorithm.The experiment results show that the proposed algorithm is more effectively by comparing with seven defogging algorithms carefully from subjective and objective aspects.The proposed algorithm needs time about one-tenth of the original algorithm,reaches quasi real-time.3.The application of the proposed algorithm in the video monitoring system.This system is mainly based on HTML5.The improved defogging algorithm is programed by OpenCV at acquisition side,defogging each frame of video.The video processed is coded and sent by FFmpeg.Meanwhile,the server uses Websocket to transfer data so that the browser can receive and play real-time video streams without plugins.