Research Of Image Haze Removal Algorithm Based On Dark Channel Prior And Retinex Theory

In recent years,there has been a large area of hazy weather in China,which has had a great impact on our normal production and life.The haze is generated by the absorption and scattering of light by particles such as a large number of microscopic water droplets,aerosols,etc.suspended in the atmosphere.It causes the contrast of the target in the image to decrease,the saturation decreases,and the color degrades,thereby making the outdoor vision system.For example,traffic monitoring at intersections,and the detection of runways by towers at airports,have reduced or even degraded results.Therefore,it is of great practical significance to clarify the smog image.The dissertation mainly studies the fog map imaging theory.Based on the atmospheric scattering model and the dark channel prior theory,this paper combines the image segmentation algorithm and the Retinex theory to propose two improved optimization algorithms.Algorithm 1: A Priori Improved Algorithm for Dark Channel Based on Sky Region SegmentationThis algorithm first studied the dark channel prior theory and found that He Kaiming's dark channel prior theory was not established in the sky region of the image,and the atmospheric light intensity was not estimated when there were large bright regions or strong light sources in the image.Be applicable.To solve these problems,the algorithm uses image segmentation to separate the bright region from the image for separate processing.The key to the method lies in the accuracy of the bright area segmentation.The algorithm for the region to the correction and atmospheric transmittance light intensity of accurate selection,thereby eliminating the classic algorithm will appear in the sky area to fog pseudocolor phenomenon and the atmospheric light intensity when the image contains strong illuminant estimation errors leading to restore image darker defects.Algorithm 2: Fast single image dehazing algorithm based on image fusionIn this algorithm,based on the atmospheric scattering model,the atmospheric light intensity is estimated in intervals;a simple estimate of the transmittance is obtained from the dark channel prior law,and a Multi-scale Gaussian convolution is used to obtain a fuzzy estimate of the transmittance by the Retinex theory,and then the image is used.The fusion combines these two pixel levels to obtain an accurate estimate of the transmittance;afterwards the cross-bilateral filtering is used for smoothing and the transmittance is corrected for the bright areas;then the final restoration can be obtained by making some adjustments to the tonality of the restored image.image.Experiments show that the algorithm not only obtainsgood defogging effect and better image color,but also effectively reduces the time complexity.This algorithm will take the threshold value adjusted semi-inverse image as the guide image.Based on pixel level,comparing the hue value for each pixel,no filtering process is used.The computation time and space complexity are greatly reduced.Moreover,the solution time of the process is small and the efficiency is improved at a large extent.The total time of the algorithm is greatly reduced,so that the algorithm can achieve real-time defogging.The average value of the first 1% brightest pixels of the original fog image are used as the reference white light compensation to further adjusts the brightness of the restored image.The restored image is clear and natural,and the contrast has been greatly improved,but the effect is not obvious for the image of the dense fog or the image taken in the night scene.