The Research And Implementation Of Video Defogging Method Based On DM8168

At present, monitoring systems have been widely used in every field of production and life. The increasing of bad weather, such as fog and haze, had made a lot of adverse effects on the quality of video captured by the monitoring system of outdoor. The image and video captured in the bad weather always are vague, which sometimes are hard to identify the scenery in them. Therefore, it has great research value to study a subsystem with the function of defogging and heightening the video. The embedded system with an open architecture is easy to operate and has strong stability as well as adaptiveness, so this paper does research on current defogging algorithms and implement these algorithms based on the DSP embedded system. The following is the main research work of this paper:Firstly, This paper mainly studies the dark channel defog algorithm and the Guide filter algorithm based on the physical model, as well as the multiscale Retinex algorithm based on the image enhancement, and optimizes three algorithm by using general methods, such as the editor optimization, inline function, removing the memory correlation and the software pipelining, and so on. Using the enhanced direct memory access implements the quick transportation of dada. In addition, the three algorithms are optimized according to their own properties.Secondly, the dark channel defog algorithm can achieve the real-time processing, by adopting the general optimization methods and down-sampling. Compared to the dark channel defog method, the guide filter algorithm mainly processing the transport map using guide filter, thus this partition manly be optimized by using frame skip and writing linear assembly files. As a result the guide filter algorithm can process video in real time and be tested in hardware platform.Thirdly, the multiscale retinex algorithm is including of multiple gauss filter, and is difficult to achieve real-time processing because of the complexity of the method. Except for the general optimization, the algorithm is improved by adopting convolution kernel optimization, the transition from time domain to airspace, the color space conversion, using function library and so on. The results show that the optimized algorithm can implement the real-time processing of video on the hardware platform and attain good defogging effects.Through the study of the three defogging methods in this paper, the three optimized algorithms achieve the real-time processing of fog video and meet the needs of most scenes. In the meantime, the optimization methods in this paper are also adaptive to the optimization and improvement of other algorithm and can avail further video processing.