结合天空区域分割和加权融合的图像去雾算法

Image Dehazing Algorithm Using Sky Region Segmentation and Weighted Fusion

针对传统暗通道去雾算法对天空区域处理能力不足,复原效果常伴有色彩失真及光晕现象等问题,提出了一种结合天空区域分割和加权融合的去雾算法。利用天空区域的亮度特性设置众数约束阈值,将雾图分割为天空和非天空区域;结合不同滤波尺寸暗通道优势,构造融合暗通道;在天空区域分割的基础上采用加权技术获得更可靠的大气光值;设置过渡区域对天空和非天空区域的透射率进行结合。实验结果表明,针对含有天空区域的雾图,本文算法的去雾效果明显,改善了天空区域颜色失真的问题,并抑制了边缘区域的光晕效应。

The traditional dehazing algorithm using a dark channel is insufficient for studying the sky area and its restoration effect is accompanied by color distortion and halo.To solve this problem,we propose a dehazing algorithm combining sky region segmentation and weighted fusion.First,the cloud map is divided into sky and nonsky regions by setting mode constraint threshold from the brightness of the sky region.Second,the fusion dark channel is constructed by combining dark channels with different filter sizes.Then,the weighted technique is used to obtain a more reliable atmospheric light value based on the sky region segmentation.Finally,a transition region is set to combine the transmittance of the sky and nonsky regions.The experimental results show that the proposed algorithm dehazes the fog map in the sky region,improves the color distortion of the sky region,and restrains the halo effect of the edge region.