Research And FPGA Design On Image Dehaze Algorithm Based On The Principle Of Dark Grey Prior

Nowadays many outdoor monitoring systems, such as road monitoring, security monitoring,etc., require high image quality, but the haze and other bad weather conditions obtained images,often because of the small particles in the atmosphere scattering effect, so that the image quality serious decline, which not only affect the image of the sensory effect, but also to the image of the subsequent processing, such as image feature extraction, image recognition and tracking of the target cause great interference. So that the outdoor monitoring system can not work properly,bringing serious production safety risks. Therefore, it is very important to clarify the image which is affected by haze and so on.At the beginning of this paper, first introduces the research background of image dehaze and the present situation of development at home and abroad. Then, the causes of haze formation are briefly described,and the atmospheric scattering model is introduced. The reason of image degradation in haze weather is analyzed, Finally locating the atmospheric scattering model based image dehaze algorithm on the principle of dark channel prior, further study of the image dehaze algorithm based on dark color transcendental principle,The principle Characteristics and defects of algorithm of the dark channel prior theory are analyzed deeply,and on this basis, the method of estimating the atmospheric light value A of the dark-channel dehaze algorithm is improved. In the original algorithm, only a value in the original fog image is taken as the final atmospheric light value. This paper uses the idea of statistical mean to improve the atmospheric light value.. Finally, the improved haze image method and other restoration methods are compared and evaluated. The original algorithm and the improved algorithm can improve the quality of the restored image on the basis of guaranteeing the running time of the original algorithm.In this paper, we use the Verilog language pair for the key algorithm in the hardware design of the image dehaze algorithm based on the dark primary a priori principle, and then use Modelsim to do the functional simulation. It is divided into three modules: the rough transmittance map estimation module, the atmospheric light value A estimation module and the coarse transmittance diagram refinement module. In the rough transmittance estimation module.two successive calculation and column minimum calculation are used to obtain the rough transmittance map. In the atmospheric light estimation module, the original algorithm is improved, and the dark channel image is used Data and the initial value to do the comparison,find the darkest channel image in the brightest pixel value corresponding to the address, and then look up the table to get the original image of the corresponding pixel image, and finally subtract it from a correction parameter as the final atmospheric light Value; in the rough transmittance diagram refinement module, a bilateral filter is used to perform the refinement operation. And finally in the Quartus Ⅱ environment was integrated, with Altera Cyclone ⅣFPGA development board was verified. The experimental results show that the output of each module is basically the same as that of board level validation.