An auto-registration method for space-borne SAR images based on FFT-shift theory and correlation analysis in multi-scale scheme

SAR co-registration is one of the crucial steps in the process of SAR image analysis, especially for multi-temporal SAR images analysis. Focusing on the repeat-pass space-borne SAR images, an auto-registration method for repeat-pass SAR complex image pair is proposed, which combines the Fourier transform and correlation-coefficient algorithm in a multi-scale scheme. Firstly, an initial offset is estimated in Fourier domain by using the 3^rd level wavelet decomposition sub-image. Then taking these initial offset into account, match-search window pairs are randomly located over reference and sensed images for correlation analysis to get the identical points which are used for estimating the registration model in fine scale. In order to evaluate the proposed method, ALOS-PALSAR and RadarSAT-2 data in different area are selected as experiment data. And the experimental results show that the proposed technique allows the co-registration of the SAR images with the accuracy up to a fraction of a pixel automatically without a priori knowledge.