摘要
图像配准是图像超分辨率重建的必要步骤,也是对重建效果影响最大的关键步骤。频域相位相关的配准算法具有抗噪声性能好和运算量小等优点,本文基于相位相关方法,提出了只对频域未受混叠影响的部分进行参数估计的配准算法。首先利用低通滤波器过滤低分辨率图像的混叠信息,然后对低分辨率图像进行傅里叶变换,再对其变换后的幅值进行极坐标变换,使其互功率谱经傅里叶反变换后为二维脉冲函数,通过二维脉冲函数的位置来得到配准参数。仿真实验表明本文方法有良好的配准精度和重建效果。
Image registration is the necessary step of super-resolution reconstruction,and also is a Key step which has the greatest effect on reconstruction.The phase correlation method in frequency domain has advantages of anti-noise performance,a small amount of computing,etc.In this paper,we put forward a registration algorithm based on phase correlation method,which estimates parameters by only aliasing-free parts in frequency domain.Firstly,we avoid aliasing parts from low-resolution image by low-pass filter,and then carry out a Fourier transform and a polar transform for its amplitude,which makes the inverse Fourier transform of the cross-power spectrum be a 2D impulse.And we can get the registration parameters from the impulse location.Simulation results show that the improvement method has good precise of registration and good effect on image reconstruction.
出处
《电子测量技术》
2010年第5期58-61,共4页
Electronic Measurement Technology
关键词
超分辨率
图像配准
相位相关
混叠
super resolution
image registration
phase correlation
aliasing
