Research And Application Of Sea-air Image Dehazing Based On DCP Algorithm

Images are the most important method for people to obtain information.However,with the evolution of the environment,the images captured by the imaging equipment contain a large amount of fog,which seriously affects the extraction and analysis of information in the images.So the field of image dehazing has been quickly developed.The main findings are as follows:1 Firstly,the development of image dehazing at home and abroad in recent years and the atmospheric scattering model are explained,and the dark channel prior DCP algorithm is introduced,and the value of at least one color channel of at least one color channel of R,G and B is approached to 0 for clear fog-free images without a large area of sky.Lay the foundation for subsequent algorithm improvements.2 Non-maritime fog-bearing images are subdivided into images without sky areas and images with sky areas.Recover depth information in images without sky areas with an improved depth model that solves an issue in the previous depth model.Improved guided filtering is used to optimize transmittance for images with sky areas,which solves the previous situation that the weight size cannot be actively changed and the image is distorted.The improved KNCM segmentation algorithm is used to divide the image into sky area and non-sky area,and the fusion atmospheric light value is calculated by combining its proportion and the selection method of the atmospheric light value of the dark channel prior algorithm,and the fusion atmospheric light value is used to correct the transmittance,which solves the problem that the lower limit of transmittance is difficult to determine before.3 Images containing fog for the sea are divided into images with a priori area and images without a priori area.The sea image with prior region is converted into the HSV model and segmented by KNCM algorithm,adaptive weighted guided filtering is used to estimate the transmittance₁ of the sky area,and the improved relative total variation model 4-RTV is used to optimize its transmittance₂ for the sea surface area,and the two are weighted and fused to obtain the final transmittance,which can solve the problem of inaccurate transmission estimation of offshore images,Compared with traditional methods,the quadtree algorithm can select atmospheric light values well.For images without prior regions,the images predicted by MMP-Net with multi-scale features multi-parallel fusion network structure can restore the image well,skip the step of calculating atmospheric light value and transmittance,greatly reduce the calculation error,replace the traditional transpose convolution operation with an improved transpose convolution operation,reduce the noise generated in the process of restoring the image,and better retain the edge information of the image.4 The system architecture of video surveillance based on image dehazing consists of image acquisition module,fog concentration detection module and image dehazing module,help people process haze-containing images and extract information.Finally,through various evaluation indicators,it is judged that the improved image dehazing algorithm in this paper can achieve improved dehazing effect in specific scenes.