提出了一种基于兴趣点不变矩(IPIM)的图像拼接技术.利用Harris角检测器获取图像中的兴趣点,计算兴趣点邻域的平移、旋转及尺度不变矩,通过比较各兴趣点邻域不变矩的欧式距离提取出初始特征点对,根据几何变换模型剔除伪特征对,最后利用正确映射模型实现图像的拼接.实验表明,该方法对平移以及任意角度的旋转具有良好的鲁棒性,对于具有小尺度变换的图像仍然具有很好的拼接效果.
An automatic image mosaic method based on invariant moments of interest points was proposed. Firstly the arris corner detector is used to gain the interest points,and the moment invariants to traslation,rotation and scale defined on the interest point neighborhood are computed. Then by comparing the Euclidean distance of these invariant moments,the initial feature points pairs are extracted,and the spurious feature points pairs are by eliminated by geometric transform model. Finally image sthtching with the correct mapping model is realized. Experimental results demonstrate the proposed algorithm is robust to translation,rotation and slight scaling.