基于尺度预估的SAR图像相位相关配准算法  

A FFT-based SAR Image Registration Algorithm Via Priority Estimation of Scale Distortion

作　　者：林岳松[1] 冯卫平[1] 陈华杰[1] 

机构地区：[1]杭州电子科技大学信息与控制研究所,杭州310018

出　　处：《现代雷达》2010年第7期39-44,共6页Modern Radar

基　　金：国防预研基金资助项目;浙江省教育厅科技计划基金资助项目(Y200702897)

摘　　要：高精度的图像配准是SAR图像融合的前提条件之一。针对相位相关匹配方法用于强相干斑噪声的多波段、多极化SAR图像匹配时存在的缺点,提出了一种基于尺度预估的SAR图像相位相关配准算法。首先,利用图像间的中心频谱值之比实现尺度预估计;然后,在相位相关过程中,针对强噪声图像的互功率谱的逆傅里叶变换图存在多峰的情况,通过脉冲值的局部搜索得到最佳的匹配参数;最后,校正待配准图像,达到精确配准。实验结果表明,文中算法对于多波段、多极化SAR图像具有良好的匹配效果。SAR image registration is a crucial step in SAR image analysis. For the shortcomings of traditional FFT-based approach for registration of multi-band and multi-polarization noisy SAR images, an improved FFT-based registration algorithm is proposed in this paper. Firstly, the ratio between central spectrum of template images and sensed-images is used to estimate the scale changes. Secondly, on the process of phase correlation, the best matching parameters are obtained by making use of the local search of pulse, overcoming multi-peaks in the IFFT of cross-power spectrum. Finally, sensed-images are corrected. It could be clearly indicated by the results that the algorithm is precise and robust for SAR images registration.

关 键 词：尺度预估 相位相关 图像配准 

分 类 号：TN957.52[电子电信—信号与信息处理]