结合小波与增量符号相关的鲁棒图像匹配

结合小波变换和增量符号相关提出了一种有效的鲁棒图像匹配方法。将实时图和基准图做小波分解以获得粗尺度的平滑图像；利用相邻像素间灰度的升降对所获得的平滑图像进行二值化重新编码；匹配二值化后的实时图和基准图，找出前者在后者中的坐标，进而确定实时图在基准图中的位置，以完成匹配。小波分解能有效的去除噪声并能大大提高匹配算法的实时性。增量符号相关方法对图像灰度值的变化不敏感。因此，二者相结合不仅有效的解决了存在噪声干扰、阴影混入（块误差）和部分遮挡等影响下的图像匹配问题，且与传统匹配方法相比，匹配速度提高了约10倍以上。理论分析说明了方法的可行性，多组实验显示了其有效性。

Robust image registration based on wavelet and increment sign correlation

By combining wavelet transform with increment sign correlation, an efficient and robust image registration method is presented. By decomposing the sensed image and the reference image with wavelet, smooth images are generated in coarser scales. By use of the increment and decrement information of adjacent pixels, the smooth images are encoded to binary images, and by matching the binary images, the registration point is found. Wavelet could remove noises effectively and could improve the registration time efficiently. Increment sign correlation is robust to small change of the image intensity. By combining them, the proposed algorithm can not only effectively match images corrupted by noises, interfused by shadows and influenced by partial occlusions, but also shorten matching time for more than 10 times compared to traditional methods. The feasibility of the algorithm is analyzed through theoretical analysis, and the effectiveness of the algorithm is illustrated by multiple experiments.