摘要
在很多应用领域,要求图像配准的精度达到亚像素级,而多数亚像素级配准方法的运算量较大。文中阐述了一种高效的基于奇异值分解(SVD)的相位相关法。在无噪声干扰的情况下,该方法的精度达到0.02像素。通过对已知平移关系的图像的配准实验,比较了其与灰度相关函数内插法在无噪声和有噪声的情况以及应用于多光谱图像配准时的配准精度。实验表明,2种方法都有较好的抗噪声性能,在受噪声干扰不严重的情况下,基于SVD的相位相关法的配准精度优于灰度相关函数内插法。
Subpixel image registration is required in many applications,but most of the subpixel image registration algorithms are unefficient.In this paper,we implement an efficient phase correlation method based on singular value decomposition（SVD）.The accuracy of this method is 0.02 pixels when there is no additive gaussian white noise.Then we compared the registration accuracy between this method and intensity correlation function interpolation method using images with varying noise level and multispectral images.Experiment results verify that the accuracy of phase correlation method based on SVD is better than correlation function interpolation method.
出处
《国外电子测量技术》
2012年第4期45-49,共5页
Foreign Electronic Measurement Technology
关键词
图像配准
亚像素
奇异值分解
相位相关法
灰度相关函数内插法
image registration
subpixel
singular value decomposition
phase correlation
intensity correlation function interpolation
