细节保持的快速曝光融合

Fast detail-preserving exposure fusion

为了获取多曝光图像序列中高动态范围的目标信息,提出一种快速的曝光融合算法.提出兼顾局部细节和全局亮度的融合权重函数.在较少时间代价的前提下,提出先下采样再上采样的权重函数计算方法,一方面更多地保留局部细节,另一方面消除像素级融合导致的光晕现象.利用导向滤波器去除上采样引入的块效应.通过10组多曝光图像序列验证算法的有效性,并将提出算法与3种典型的曝光融合算法进行对比.实验对比结果表明:该算法具有最强的细节保持能力,可以同时兼顾目标场景的全局亮度分布.量化对比结果表明:该算法可以获得较好的融合效果并且时间代价较小.

A fast exposure fusion method was proposed to obtain high dynamic range target information from multi-exposure sequence. A new weight map was proposed, which gave consideration to both local details and global brightness. Secondly, on the premise of small time cost, a down-up-sampling computing method for the weight map was proposed, which could preserve local details well and eliminate artificial halos caused by pixel level fusion. And then, guided filter was used to remove the block artifact caused by up-sampling. The validity of the proposed fusion method was demonstrated by ten multi-exposure sequences, and the proposed fusion method was contrasted with three typicals of fusion algorithms. The experimental results show that the proposed fusion method has strongest retention capacity of the local details and the global brightness distribution of target scene. The quantitative comparison results show that the proposed method achieves better fusion qualiity, meanwhile, and the time cost is pretty small.