摘要
文中针对基于梯度算子的图像匹配方法通常存在的问题———收敛速度较慢、迭代易落入局部最优点，提出了一种基于小波变换的多尺度图像匹配算法。这一算法利用小波的良好的多尺度特征，先对低分辨率的子图像做匹配，将得到的结果作为初始值用于高分辨率图像的匹配运算，匹配运算都在位移矢量场上进行。实验表明该方法可节省计算时间，且减小迭代落入局部最优点的概率，算法鲁棒性得以提高。
In this paper, hierarchical image matching algorithm based on wavelet is presented. Due to the multiresolution feature of wavelet representation, the advantages of this algorithm are that the convergent rate is faster and the results are not so easy falling to the local minimum. Our algorithm matches subimages in the lower resolution level to get the displacement vector field between two subimages, then this matching result is passed onto the higher resolution level as an initial estimate. Experiment proves that the algorithm indeed accelerates the convergence rate, and the probability that the solution falling into the local minimum is decreased. \;
出处
《红外与激光工程》
EI
CSCD
1999年第4期30-33,共4页
Infrared and Laser Engineering
关键词
图像匹配
小波变换
多尺度
位移矢量场
Image matching\ \ Wavelet transform\ \ Multiresolution\ \ Displacement vector field
