基于Fourier-Mellin变换和曲面拟合的IC芯片图像亚像素配准算法

IC Chip Image Sub-pixel Registration Algorithm Based on Fourier-Mellin Transform and Surface Fitting

为实现IC芯片图像的精确配准,提出一种Fourier-Mellin变换和曲面拟合法结合的图像亚像素配准算法。算法首先采用Fourier-Mellin变换计算出待配准IC芯片图像的旋转角和平移量的整像素峰值位置,然后对峰值区域进行基于最小二乘的曲面拟合,通过取极值点获得亚像素级的精确峰值位置。另外,Fourier-Mellin变换由于旋转的频谱混叠和坐标变换中的重采样插值会产生误差,新算法采用加窗和滤波的方法加以改进。实验结果表明,所提算法实现了IC芯片图像的亚像素级配准,具有较好的精度和鲁棒性。

To achieve IC chip image accurate registration,a new image sub-pixel registration algorithm which combines Fourier-Mellin transform and surface fitting method is proposed.The new algorithm first calculates the peak position of rotation angle and translation of the matched IC chip image via using Fourier-Mellin transform.Then,the peak area is fitted by using least squares method,and the accurate peak position which is a sub-pixel value is obtained through calculate extreme points.In addition,spectrum aliasing and resampling interpolation during coordinate transformation will bring errors by applying Fourier-Mellin transform.The proposed algorithm resolves this problem by adding window function and filtering.Simulation results show that the proposed algorithm accomplished IC chip image sub-pixel registration with optimal robust and accuracy.