Research On The Algorithm Of Fog Removal In Foggy Image Of Water Region Based On Physical Model

Under haze weather conditions,optical imaging systems usually can not obtain information normally,which greatly affects the performance of the optical imaging system.At the same time,the environment with rivers,lakes and sea surface is more likely to produce haze and other bad weather because of the more abundant water vapor,which makes the imaging system more challenging to gather information in water area.Therefore,the in-depth study of the fog image algorithm in water area is in line with the urgent need of practical application and has wider application value.In view of the fact that the visibility and contrast of the foggy images in the water area are reduced,the details of the scenery are blurred and the color is distorted.Based on the two directions of the atmospheric single scattering model and the atmospheric multi scattering model,we studied the removal algorithm of water area image based on physical model.The main research work involved in the following aspects:Firstly,the algorithm principle analysis of the two main de-fogging algorithm He algorithms and the Tarel algorithm based on the atmospheric single-scattering physical model is analyzed.The applicability of these two algorithms to the de-fogging of watershed images is analyzed based on the characteristics of the watershed scenario.In this paper,we propose a fast filter de-fogging algorithm based on dark channel a priori segmentation region and an improved de-fogging algorithm based on atmospheric veil.A higher accurate estimation method of atmospheric light is proposed in the fast filtering algorithm based on the dark channel priori segmentation region.The halo effect is avoided by the segmented region filtering,while the speed of processing is faster compared to the He algorithm using the guided filter.The improved algorithm based on atmospheric mask is to estimate the air mask with the scene information under the condition of following the atmospheric mask.In order to avoid the fog distortion in the sky area,the estimated atmospheric mask is adjusted according to the similarity between the pixel and the atmospheric light.For the fog removal results,the hue mapping algorithm which is more suitable for the water scene is adopted.Compared with the Tarel algorithm,the improved atmospheric mask algorithm is more suitable for the fog removal in the water area.Then,combined with the characteristics of the atmospheric multiple scattering model to remove the fog clearer and the dense fog of the water area is more frequent,this paper proposes a fog removal algorithm based on the atmospheric multiple scattering model,which mainly includes two parts: the more accurate transmittance map is obtained and the deconvolution is recovered.In order to solve the problem of inaccurate transmittance map of fog removal algorithm based on atmospheric multi-scattering model,this paper proposed a local search method based on dark channel a priori similar pixels,which can obtain more accurate rough transmittance map,While avoiding large areas of white area transmittance error estimation.Aiming at the existing problems of the deconvolution of the convolution kernel based on the decontamination algorithm based on the atmospheric multi-scattering model,the convolution kernel mapped to the image domain is more reasonable by selecting the appropriate mapping template.In the whole figure,the deconvolution calculation is proposed,and the ambiguity caused by other causes is solved while resolving the ambiguity caused by the multi-scatter phenomenon in the atmosphere.This method can improve the clarity of the fog results,and is more suitable for the use of fog removal in the water fog.Finally,the objective evaluation method of image quality is introduced.Combined with the image quality evaluation system based on visual perception and the index of energy gradient function,we objectively evaluate the fog removal results of water scene in this paper.The evaluation results show that the algorithm proposed in this paper has better effect on fog removal in water area image,and has obvious advantages in improving image definition and recovering more scenery details,and it can effectively improve the quality of water fog image.