Research On Real-time Video Defogging System Based On FPGA

In foggy weather,the scattering of atmospheric particles will greatly reduce the quality of the image obtained,and the details of the image will be seriously covered up.The decline of image quality will make it difficult to carry out the work of recognition and tracking,so the research of video image defogging is of great significance.At present,the defogging algorithm is still mainly implemented on PC,so the processing speed is greatly limited.On the basis of the characteristics of high parallelism,small size and low power consumption of the FPGA,we designed a real-time video defogging system with the core of the FPGA in this paper.Firstly,we studied the image defogging algorithm in depth and analyze the hardware realizability.We compare the enhancement effects of each algorithm through the implementation based on OpenCV.Then we chose CLAHE algorithm and dark channel prior algorithm to optimize and implement on FPGA.We built a video system on the hardware platform with Sparten 6 series of FPGA chips as the core.And we completed video image acquisition,DDR cache and 1080 P VGA output,which provided a reliable video source for subsequent image processing systems.On the basis of deep understanding of the algorithms,we propose improvements to the realization of the two algorithms in view of the design objectives:(1)For CLAHE algorithm,we set the shear ratio of each sub-block adaptively,and propose a scheme to calculate the shear parameters with the average brightness and contrast values of the image.It makes the local processing result better and adapts to various scenarios.We optimize the histogram reallocation process and improve the efficiency of the algorithm.And we improve the execution process of the algorithm to reduce the computational complexity by shifting and counting.(2)For the dark channel prior algorithm,we propose a scheme to compensate the transmittance.It solves the problem of poor processing effect in bright areas such as sky without additional delay and complex operations.For the estimation of atmospheric light value,we use histogram statistics to avoid sorting in the original algorithm.This design saved a lot of resources and improves the efficiency of the algorithm.We use the equivalent transformation of the formula to optimizing the calculation method of transmittance,which reduces the system delay.Finally,we implement two defogging algorithms based on FPGA.The theoretical analysis and experiment prove that the system has good defogging effect.Both delays of two algorithms are in microsecond order,which meets the real-time requirement.The system can efficiently process 1920×1080@60Hz video images.The optimized enhancement algorithms have practical value and are also suitable for application in the field of medical and aerospace.