一种快速暗通道去雾算法

A Fast Dehazing Algorithm Based on Dark Channel Prior

摘要:

基于暗通道的去雾算法计算初始透射率需要进行大量的数据比较,优化透射率时需要计算融合矩阵,这两个过程耗时巨大,使其难以投入实际应用。针对这一问题,提出了一种快速暗通道去雾方法。首先利用分区最小表的数据结构,提高初始透射率的计算速度；接着,在优化透射率时,采用基于形态学梯度的彩色图像边缘检测算法提取图像边缘信息,减少优化范围,再利用边缘信息与像素的空间信息优化透射率,避免计算复杂矩阵,从而加快了优化速度；最后,运用最小可觉差模型补偿前两个步骤导致的图像画质下降,提高图像清晰度。实验证明,快速暗通道去雾算法在保持恢复图像效果基本不变的基础上,很大程度提高了算法速度。

Abstract:

The dehazing algorithm based on dark channel prior is hard to be put into practical application due to the great time complexity of estimating and optimizing the transmission, which needs a large amount of data comparisons and calculating the amalgamation matrix. In order to solve this problem, an fast algorithm based on dark channel prior is presented. Firstly, a data structure called partitioned minimal table(PAMT) is used to increase the estimating speed of initial transmission. Then, a color image edge detection based on morphological gradient is adopted to obtain the edge information of the image to reduce the scope when optimizing the transmission. This edge information and the spatial correlation of the pixel used to optimize the transmission can avoid the complex matrix calculation, which accelerates the speed of optimization. At last, the final recovered image, with enhanced contrast, is obtained by performing a post-processing technique based on just-noticeable difference(JND). Experimental results show that the performance of the proposed algorithm is substantially the same as the original one, but the time complexity has greatly reduced.