Research On Real-time Video Dehazing Algorithm Based On FPGA

Harsh weather such as rain and fog make atmospheric scattering enhanced,which leads to serious degradation of the video images collected by outdoor imaging equipment,and the application of actual scenes is greatly limited.Therefore,it is very important to study better defogging technology based on image enhancement or image restoration methods to remove the influence of weather factors on image quality.The defogging method based on image restoration,inverts the image imaging process,and then use the atmospheric degradation model for defogging,which has achieved good results,and thus has become a research hotspot.However,the performance of the existing image restoration-based defogging method needs to be improved,and it has some limitations when transplanted to the hardware platform,the defogging task cannot be completed well.Aiming at this problem,this paper introduces an efficient adaptive threshold dehazing algorithm based on dark channel prior and edge preservation,which reduces the complexity of the defogging algorithm and improves its hardware portability.Combined with the FPGA "pipeline" structure and parallel data processing,a dehazing algorithm based on dark channel is proposed to meet the real-time processing requirements of video images.The main research contents of this thesis are as follows:(1)For the limitations of dark channel prior defogging algorithm when dealing with images containing much sky regions,high complexity and resource consumption when the device is transplanted with hardware.The adaptive threshold dehazing algorithm for dark channel prior and edge preservation is proposed to introduce Sobel operator to refine the transmittance t,and use the adaptive threshold to limit the atmospheric light value A to reduce the algorithm complexity.Secondly,from the hardware level,the new algorithm weakens the correlation between A and t,and the parallel acceleration is performed.The algorithm of this paper simplifies the process of obtaining key parameters while improving the image defogging effect,and improves the hardware portability of the algorithm.Finally,comparing and analyzing the experimental results by the objective evaluation criteria of the image.The algorithm provides a theoretical basis for transplanting to the FPGA hardware platform.(2)Propose and implement an FPGA-based real-time video dehazing algorithm.An FPGA image processing platform integrating video image acquisition,storage and display is designed,which realizes miniaturization of the defogging device and provides a hardware platform for algorithm implementation.After decoupling the adaptive threshold dehazing algorithm based on dark channel prior and edge preservation,the algorithm is modeled by modular design idea to transplant the algorithm to the FPGA hardware platform.The paper shows simulation results of image edge retention module,dark channel acquisition module,atmospheric light value seeking module,transmittance seeking module and other important parts of the module,to verify the rationality and accuracy of the algorithm,and finally realize real-time recovery of foggy images.Comparing the Matlab simulation results,the optimized algorithm improves the clarity of the defogging effect.Model Sim simulation results verify the enforceability and correctness of the algorithm after the module is divided.The board level test results show the image after defogging.The system processing speed can reach 30 frames/s,which meets the needs of real-time processing of video.