The Algorithm Research And Implementation Of Image Dehazing In Video Image Based On FPGA

In recent years,the image processing has been developed so rapidly that it has been used widely in people's daily life.But the quality of the collected images are deeply affected by external environment.For example,there are some problems such as blurriness and dark-oriented brightness existing in the collected foggy images,so the visual result is affected and it is difficult to get valuable information.At present,in the field of target recognition and image tracking,the problem of how to quickly and efficiently defog and improve the quality of the image needs to be urgently solved.In this paper,two de-fog algorithms are described in detail.On the basis of algorithm of dark channel prior(hereafter will be abbreviated as He algorithm),the paper puts forward improved algorithm.Through the deep research in He algorithm,He algorithm has two deficiencies in dealing with video image:1,He Algorithm can not apply to region of the sky and the bright area is likely to be distorted.2,It will appear halo if using a rough transmittance to recover brilliant image,it's also too complicated and time consuming by using soft matting(Guided Filtering)to refine the transmittance.So it is difficult to realize in hardware platform and unable to meet the requirements of real-time defogging.Against the above two deficiencies,considering the feasibility on the FPGA hardware platform and the premise of the real-time video processing,the paper puts forward the following methods: 1,setting adaptive threshold to segment the image,then reducing transmittance's deviation by adjusting transmission to eliminate color distortion.2,As the process of refining transmittance in He algorithm is too complicated and time consuming,the paper improves the refining process.The purpose of refinement transmittance based on the minimum filter to inhibit the halo is achieved through refining dark channel.The second algorithm is fast de-fog algorithm which estimates the spreading images by using the luminance component,finally recovering the original explicit images based on the atmospheric scattering model.It is simple and reliable as well as easy to achieve,it also can deal with high resolution image.Then using the Altium Designer is to devise the hardware circuit and to accomplish testing hardware platform.Finally the improved algorithm transplantation and optimization on the FPGA hardware platform is completed and defogging in video image can be realized.The results tested by MATLAB simulation and hardware system indicate that two de-fog algorithms can better remove the fog in video image on the premise of outputting a high qualified video image.In addition,with a simple design and a low cost,the system has a very good practical value.The system can also be connected to other systems based on the actual need,such as providing the high quality video source for the license plate detection and recognition.