针对明亮区域的自适应全局暗原色先验去雾

针对暗原色先验去雾算法对明亮区域失效,以及分块求取暗原色存在的块状效应、Halo现象和运算复杂度较高等问题,提出了一种基于自适应参数的全局暗原色先验去雾算法。该算法采用全局暗原色操作取代分块处理,并通过模糊逻辑控制器自适应估计明亮区域的容差参数和透射率调整因子;在非明亮区域求取大气光强度后,根据自适应容差纠正明亮区域被错误估计的透射率。与常用的3种图像复原去雾算法进行了比较,结果表明:该算法去雾图像的主观视觉效果较好,且图像对比度、信息熵和平均梯度3方面的客观评价结果也明显优于其它3种对比算法。该算法可有效解决明亮区域失真和分块处理带来的上述问题,在不增加曝光处理情况下也能获得较好的去雾效果,运算效率也有较大提升。

Adaptive image dehazing for bright areas based on global dark channel prior

An adaptive image dehazing algorithm based on global dark channel prior was proposed to solve the invalidation of original dark channel prior algorithm in bright areas and problems of block effect, Halo effect and higher computational complexity. In this method, the blocking operation was substituted by a global dark channel operation, and the fuzzy logic controller was used to estimate adaptively the threshold of bright areas and the adjustment factor of transmission. After the atmospheric light was estimated in non-bright areas, the miscalculated transmission in bright areas was corrected according to the adaptive tolerance. The algorithm was compared with three kinds of image restoration dehazing algorithms. Experiment results show that the algorithm shows a good subjective visual effect for dehazing images, and the objective evaluation criteria, image contrast, information entropy and average gradient are also superior in performance to those of the other algorithms compared. It concludes that the presented method effectively eliminates the distortion in bright areas and solve the above problems caused by blocking, and the visibility of dehazing image and the operating efficiency have been enhanced significantly.