AN IMAGE REGISTRATION METHOD BASED ON THE COMBINATION OF MULTIPLE IMAGE FEATURES

The image registration is one of the key steps to achieve three-dimensional (3D) localization and the other image fusion processes. This article presents a registration method based on the combination of edge feature and corner feature. The processing steps include image segmentation, corner detection, edge detection, extraction of interested region, and correspondence points matching. The algorithm flowchart and implementation of every step are presented. The proposed method makes the image registration with the multiple features of two images and avoids the inaccuracy of the SAR image information. Therefore, it has a better robustness and registration precision. A TerraSAR-X SAR image and a GeoEye-1 optical image of the Water Cube in Beijing are processed and the results validate the effectiveness of the proposed method.