基于暗元先验的去雾算法在利用边缘保护操作消除景深突变处产生的光晕现象时没有区分边缘的类型, 导致了透射率的估计不合理, 降低了去雾的质量。提出一种基于加权融合策略的透射率估计方法, 通过块级暗通道和像素级暗通道的相关特性获取景深信息导向图, 从而在景深突变处和非景深突变的局部合理选取像素级暗通道和块级暗通道, 保护了景深边缘处的突变性, 同时减少了局部纹理边缘噪声的影响, 获得了更准确的透射率估计结果。实验结果表明, 该算法能有效避免光晕现象, 改善局部细节模糊的问题, 得到更优的去雾视觉效果。

The haze removal algorithm based on dark channel prior uses edge protection operation to eliminate the halo artifacts effect caused by abrupt change at depth edges, which is unable to distinguish types of edges, making the estimated transmission map inaccurate,and reducing the quality of dehazing. A method via weighted fusion strategy was proposed to estimate transmission, through the property of pixel-wise dark channel and block-wise dark channel to obtain the depth information guided map,thereby the pixel-wise dark channel and block-wise dark channel were selected reasonably at edges of depth abrupt change and the rest regions. It protected the discontinuity of depth edges, reduced the impact of local texture edge noise and acquired more accurate estimation of transmission. Experimental results show that the algorithm can effectively avoid the phenomenon of the halo, improve the fuzziness of local details and acquire better visual effect of dehazing.