针对多曝光的拉普拉斯金字塔融合算法存在部分过曝光以及暗部细节丢失问题, 用相位一致性滤波函数替换原算法融合系数中的对比度函数, 改进算法在提取边缘时可以忽略光照和对比度变化对图像的影响, 使得融合结果更自然, 细节更丰富.通过多组序列图像验证了算法的优点, 并与原算法进行比较.结果表明, 主观评价方面, 该改进算法的融合结果自然, 在强光背景下能较好地保留暗部细节, 减少光源处的光晕; 客观评价方面, 该算法融合的高动态范围图像的方差和直方图结果明显优于原算法, 所保留的信息多于原算法.

Aiming at the partial overexposure and the loss of shadow detail in the multi-exposure Laplacian pyramid fusion algorithm, to replace the contrast function in the fusion coefficient of the original algorithm by using the phase congruency filter function was proposed. The algorithm can ignore the illumination and contrast when extracting the edge. The effect of the change on the image makes the fusion result more natural and richer in detail. The advantages of the algorithm were verified by multiple sets of sequence images, and compared with the original algorithm. The results show that in terms of subjective evaluation the fusion result of improved algorithm is natural, and it can better preserve the details of the dark part under the strong light background and reduce the halo at the light source. In terms of objective evaluation, the variance and histogram results of the high dynamic range image fused by the algorithm are obviously better than the original algorithm, and more information is retained than the original algorithm.