Abstract: In order to solve the poor matching problem of SIFT algorithm in uneven illumination, an SIFT matching algorithm（MSHFHS+SIFT） which is fused with difference of homomorphic filtering is proposed in this paper. In the proposed method, first, the homomorphic filtering algorithm is extended to multi-scale space to build a pyramid of homomorphic filtering. Then, it fuses with SIFT algorithm for using its  feature points extraction and descriptor generation method. Finally, it uses distance-radio for matching. Experimental results in various illumination changing environments show that the proposed algorithm has better robustness than the traditional methods.

Key words: Scale Invariant Feature Transform（SIFT）, uneven illumination, homomorphic filtering, multi-scale space, image matching

摘要： 为了解决SIFT算法在光照不均环境下匹配效果差的问题，提出了一种融合差分同态滤波算法的SIFT匹配算法（MSHFHS+SIFT算法）。该方法对同态滤波算法进行扩展，推广到多尺度空间，以构建同态滤波金字塔，与SIFT进行融合，利用SIFT算法的特征点提取、描述子生成方法，根据distance-radio准则进行匹配。在各种光照变化环境下的实验证明，该算法较之传统的方法具有更好的鲁棒性。

关键词: 尺度不变特征变换（SIFT）, 光照不均, 同态滤波, 多尺度空间, 图像匹配