The Research And Application Of Single Image Defogging Algorithm

Under the haze,a large number of aerosol particles are suspended in atmosphere,the scene light will scatter with those particles during the transmission and the energy is attenuated,resulting in the quality of the collected image is impaired,which seriously affects the image feature extraction and has a serious impact on the fields relying.on image processing technology such as transportation,military and aviation.There fore,how to restore the original appearance of foggy image by defogging in order to improve the image quality is a research project with great practical significance.This paper mainly studies the defogging algorithm based on the atmospheric scattering model.It first introduces the atmospheric scattering model,which essentially analyze the degradation mechanism of foggy image.Then the dark channel prior defogging algorithm that play an important role in the image defogging algorithm is introduced,the principle and the inadequacy that need to be improved of this algorithm are analyzed in depth.In view of those inadequacy,the improvements made in this paper are divided into two parts-algorithm accuracy and algorithm efficiency.The former is subdivided into the following four steps:First,the minimum value filter is limited by a threshold to obtain a restricted dark channel map which can solve the white halo phenomenon appearing in the abrupt change of image depth;Secondly,this paper uses the quadtree subspace hierarchical search method to obtain the atmospheric light value,which effectively avoids erroneously selecting values from the white scenes in the image;Thirdly,constructs the membership function of the bright area,which is used to modify the atmospheric transmittance of the bright area to avoid color distortion;Fourthly,performs gamma transformation on the restored image to improve the overall brightness and enhance the visual effect.The latter proposes two fast edge-preserving filtering algorithms by studying edge-preserving image filtering algorithms—fast edge-preserving filtering based on variance and adaptive exponentially weighted moving average filtering,replacing the soft matting algorithm in the original algorithm to improve defogging algorithm efficiency.Finally,in order to verify the effectiveness of the defogging algorithm in this paper,using this algorithm and four other typical image defogging algorithms to defog the five foggy images in different scenes through experiments,meanwhile,making a comprehensive comparison of the experimental results from three aspects:subjective evaluation,objective evaluation,time efficiency.The experimental results show that the algorithm in this paper is more robust,not only has a good effect,but also has a significant improvement in execution speed.In addition,this paper applies the defogging algorithm to license plate recognition in haze which successfully improves the accuracy of license plate recognition.