摘要
提出了一种基于小波分解和混合环形投影特征的具有旋转不变性的图像匹配算法。首先在小波变换后的低频图上利用混合环形特征相似性算法得到一系列候选匹配点,然后对候选点在原图像对应的区域内作精确相关匹配得出准确匹配位置。混合环形投影特征值结合了均值投影和方差投影的优点,能有效反映图像的旋转不变性特征。实验证明该算法保持匹配精度的同时减少了匹配计算量,而且对图像的旋转不敏感。
An invariant-rotation image matching algorithm was proposed based on wavelet-decomposition approach and hybrid ring-projection feature. After wavelet-decomposition, similarity detection of hybrid ring-projection feature was applied to the low frequency part of the image to obtain a series of candidate matching points, and then fine matching operation was performed to corresponding ones in the original image. The hybrid ring projection value combined the mean ring-projection method and variance ring-projection method; it can reflect the feature of rotation-invariant effectively. The experimental results proved that this method significantly reduced the computational burden and was rotation-invariant, while maintaining high precision of the match.
出处
《顺德职业技术学院学报》
2012年第3期1-4,共4页
Journal of Shunde Polytechnic
关键词
模板匹配
小波分解
混合环形投影
template matching
wavelet decomposition
hybrid ring projection
