基于暗通道先验的图像去雾算法改进

Improved Image Haze Removal Algorithm Using Dark Channel Prior

暗通道去雾算法存在导向滤波精细化求取透射率后边缘细节不突出和剧烈变化的边缘处有伪影,针对这种问题提出一种基于优化的Sobel算子对导向滤波器进行自适应加权改进;同时通过判断天空区域是否存在并结合约束条件法解算大气参数,实现透射率的补偿,从而解决高亮区域失真的问题。实验结果表明,与采用统一的规整化因子导向滤波器的传统暗通道算法相比,本文算法去雾后结果图像边缘更突出,更清晰,同时消除高亮区域失真。

Dark channel defog algorithm exists guided filter refinement to obtain transmittance after the edge of the details are not prominent and dramatic changes in the edge of the artifacts And the Sobel operator based on optimization is used to improve the adaptive weighting of the guide filter. At the same time,the global atmospheric parameters are solved by determining whether the sky region exists and combined with the constraint condition method,so that the transmittance is compensated. Highlight the problem of regional distortion. Experimental results show that the edge of the proposed algorithm is more prominent and clearer than that of the traditional dark channel algorithm combined with the sky region,and it can eliminate the highlight region distortions when compared with the conventional filter method with uniform regularization factor.