Research On Image Defogging Algorithm Based On Logarithmic Attenuation Dark Channel And Fusion

Scattered and refracted by aerosol molecules of various sizes in the near-earth atmosphere,the light—combined with certain temperature and humidity conditions— can lead to fog,haze and other complex weather,which may result in the serious degradation of initial images acquired by the computer vision system,and even low contrast.Thus,it is hard to collect enough information for advanced tasks such as feature analysis and target tracking,which has adverse effects on the normal operation of mineral resources exploration,aviation navigation and other systems.Therefore,the relevant technologies of image inpainting and defogging can demonstrate the research significance and extensive practical value.According to the in-depth research of the advantages and disadvantages of existing various defogging algorithms,based on atmospheric scattering model,the following two new image restoration defogging algorithms are proposed in this paper:Algorithm I: Image fusion defogging algorithm based on minimum channel and logarithmic decrementAiming at the problem of residual fog at the edge after Class I defogging algorithm of dark channel,a fusion defogging method —based on minimum channel and logarithmic decrement—is proposed in this paper: Firstly,the minimum channel graph of foggy image is taken for logarithmic decrement to “E” at the bottom to acquire logarithmic decrement dark channel.And the transmittance is roughly estimated by using the prior assumption that the logarithmic decrement dark channel approaches zero,then the joint bilateral filtering is carried out to refine the transmittance and eliminate the texture effect,with t he guide graph as the minimum channel graph.Downsampling operation is done before filtering,and upsampling operation is done after filtering to improve the efficiency of algorithm and obtain the initial transmittance.Secondly,the logarithmic decrement is made for the edge acquired by detecting the minimum channel graph with Canny operator to obtain the edge information graph,with the initial transmittance and the edge information graph weighted and fused to form the optimized transmittance.Finally,the atmospheric scattering model is inversely solved by combining the atmospheric light value obtained by the improved quadtree search algorithm to restore the fog-free image.Consequently,the experimental results show that the algorithms in this paper can effectively suppress halo formation,remove residual fog on the edge,and have good real-time performance.Algorithm II: Fast image defogging algorithm based on logarithmic mapping and adaptive toleranceA fast defogging method based on logarithmic mapping and adaptive tolerance is proposed,aiming at the problem of sky color distort ion in existing image defogging algorithms.Firstly,a rough estimate of transmittance is acquired by using the minimum channel graph of foggy images,and a logarithmic mapping is performed to correct the coarse estimated transmittance.Secondly,a new adaptive tolerance mechanism is designed to adjust the transmittance value of bright areas such as the sky according to the proportion of the number of pixels with pixel gray value greater than 170 to the total number of pixels.Finally,the optimum transmittance obtained by the L1 regularization and the atmospheric light value obtained by the local atmospheric light strategy are used to solve the atmospheric scattering model and restore fog-free images.With reference to the experimental results,the proposed algorithms have low time complexity compared with the comparison algorithm,with the sky area restored naturally and residual fog on the edge effectively removed.