Research On Single Image Dehazing Algorithm Based On Dark Channel Prior

In the fields of intelligent driving,surveillance,and others,high-quality images are required for subsequent image processing.However,images captured in foggy conditions often have poor quality,which can have a serious impact on the results of image processing.Therefore,research on image dehazing technology has important practical significance.This article proposes solutions to the halo effect and sky region failure issues in traditional dehazing algorithms for dark channel prior.The main contributions of this article are as follows:(1)To address the problem that the dark channel prior dehazing algorithm has a halo effect in the region of the sudden change of scene depth,an improved dark channel dehazing algorithm combining haze-line prior is proposed.Firstly,the atmospheric light solution of the dark channel dehazing algorithm is improved.The characteristics of the locations where the atmospheric light is located are first analyzed,and then the candidate atmospheric light points are clustered to select the class cluster that best matches that characteristic,and the atmospheric light value is solved in that region to avoid the influence of the white scene on the solution of the atmospheric light value.Subsequently,employing the dark channel algorithm to calculate the initial transmittance,and based on the principle of haze-line prior,proposing the assumption that the transmittance of pixel points within the same quantization range in a hazy image remains consistent.Utilizing this presumption,refining the initial transmittance,the issue of halo effect is solved.Then,the adaptive tolerance mechanism is used to improve the dehazing effect of the sky area.Ultimately,converting the restored image to the HSV space allows for luminance compensation.(2)For the problem that the dark channel prior theory fails in the sky region,the reasons for its failure are analyzed,and a sky-dehazing algorithm based on region segmentation and the cost function is proposed.Firstly,the gradient map of the fog map is obtained by edge detection,and the boundary between the sky and other scenes is obtained by finding the maximum connected area for the gradient map,and the local OTSU algorithm is performed at this boundary to obtain the accurate sky segmentation results.The dehazing process is completed in the sky region using the cost function-based dehazing method for this region,and finally the dehazing result of the sky region is fused with the dehazing result of the non-sky scene to obtain the complete dehazed image.(3)The algorithm of this paper was compared with other classical algorithms to verify the effectiveness of the proposed algorithm from both subjective and objective perspectives.The experimental results show that the algorithm of this paper can effectively improve the problems of the halo effect and sky area failure in the dark channel algorithm,and the dehazing effect is clear and natural,showing certain advantages in both subjective and objective evaluations.