基于特征点匹配的拼接算法研究与实现

针对图像问因具有旋转及光线强度差异等现象而导致的拼接效果不佳及拼接速度慢的问题，提出一种基于特征点的配准算法。该算法首先利用相位相关法确定图像重叠区域，然后采用改进Harris角点检测算法检测角点，再根据相似测度NCC（NormalizedCrossCorrelation）方法提取出匹配特征点对，最后用渐进渐出的方法实现拼接图像的融合。实验证明，该算法比传统算法在角点数目上减少了四分之一，有效地去除了拼接产生的鬼影现象。能有效地提高图像拼接的速度和精度。

Research and Implementation on Image Mosaic Based on Feature Points Matching

In order to solve the problem of poor effect in image stitching caused by the rotation, and the differ- ences in light intensity, an image stitching algorithm based on the feature points is proposed in this paper. The algorithm first determines the overlapped area in images using the method of phase correlation, then extracts the comers with improved Harris detecting algorithm, and the feature point matching pairs acquired by NCC （nor- malized cross correlation） method, finally we are able to complement the fusion of stitching images by using the step-in-step-out method. The experiment results show that the algorithm improves the speed and accuracy of im- age stitching effectively.