Research And Implementation Of Real-time Video Dehazing System Based On FPGA

Recently, with the development of industrial technology, environmental deterioration and fog weather has become a severe problem and caused serious impact on social production and daily life. Image dehazing, as an important part of image pre-procesing, has a great impact on image post-processing and computer vision. Therefore, image dehazing is becoming a hot research topic. With the rapid development of computer and semiconductor technologies, small HD real-time video processing system is in demand. Therefore, it’s important to study video dehazing system with HD resolution and real-time feature.Firstly, this thesis gives a brief introduction of dehazing method and its research status home and abroad. An overall design architecture of the video dehazing system is given. A detailed comparison and demonstration is taken for each subsystem, including video input and output, video frame buffer and internal bus interface. Then, circuit and logic framework of the video dehazing system is presented.Secondly, the principle of single image dehazing method based on dark channel prior is outlined in detail. The computational complexity of the dehazing method is dicussed and the improved dark channel computation algorithm is given. The overall computational time has been greatly reduced while keeping good dehazing effect.Thirdly, the implementation of dark channel image module, fog transmittance image module and hazed image recovery module on FPGA is given with the utilization of Xilinx Vivao HLS tool. According to time-domain continuity characteristics of video, auxiliary video dehazing method is described to solve video flickering problem. Software and hardware co-processing implementation of the method on FPGA is addressed.Subsequently, an analysis of the external frame buffer memory interface, dehazing module, video I/O module and internal bus interface is carried out and performance bottlenecks of high delay and low throughput is examined. The transmission delay and overall bandwidth optimization is described to support 720 P HD real-time video dehazing processing.Finally, RTL verification of the image dehazing method and system functional test are performed. The HD video dehazing effect comparision chart is given. The test results shows that the video dehazing system achieves the expected real-time HD video dehazing function with good dehazing effect and self-adaptive feature.