基于多尺度增量符号相关的图像匹配算法

Image matching algorithm based on multi-scale increment sign correlation

将影像多尺度分解思想和增量符号相关方法相结合,提出了一种图像匹配算法。首先使用非线性的关键点滤波算子将参考图和待匹配图分解为不同尺度的子图;然后采用增量符号相关方法对影像进行二值化编码并进行相似性度量;最后通过由粗至精的逐层匹配得到最终的匹配结果。实验结果表明:在匹配区域存在一定灰度反差的条件下,该算法有效地解决了噪声干扰、光照差异大和阴影遮挡等条件影响下的匹配问题,同时大大的缩短搜索时间,提高了匹配定位速度。

An image matching algorithm based on multi-scale increment sign correlation is proposed in this paper. Firstly, images were decomposed to different scale-space sub-images with rich pixel features by nonlinear critical point filter. Then, according to the increment and decrement information of adjacent pixels, the sub-images were encoded to binary array. Their correlation similarities were measured by increment sign correlation. Finally, image matching was performed from global to level-wise matching with the intuition as prior knowledge. The experimental results show that this algorithm is of computational efficiency and that it is robust under gray change or partial occlusion conditions with small pixel＇s value difference in matching area.