A Multi-Step Strategy for SAR Image Matching

SAR Image matching is a critical step in the radargrammetry, interferometry, DSM Generation and change detection applications of SAR image data. Image matching procedure in these images is much more difficult than the optical images due to high speckle noise and geometric distortions defining by layover, shadow, and foreshortening. Therefore, study on an efficient SAR image matching could be useful. In this paper, a multi-step strategy for this issue is proposed. These steps include SAR image despeckling, point feature extraction, feature allocation, local image matching and global image matching. In the proposed multi-step method we used current algorithms in photogrammetry and computer vision. In each step, several algorithms are experimented and compared to specify the final algorithm for that step. In our experiments, we used a pair of TerraSAR-X single-look slant-range complex (SSC) images with short and long baselines that were acquired over the city of Jam, southern Iran, in spotlight mode. The result shows that the proposed method could match appropriate number of points in both images with high accuracy and reliability. For example it could match 211 points with 1.9 pixel accuracy for long base line image pair with size of 700×700 pixels and 1603 points with 1.22 pixel accuracy for short base line image pair with size of 400×400 pixels. Therefore, the proposed SAR multi-step image matching strategy is appropriate for coarse matching level.