Registration of RapidEye and TerraSAR-X Satellite Images Using Invariant Local Features and Non-rigid Geometric Transformations

Abstract

In this paper, a feature-based method for registering Rapid-Eye and Pleiades optical images with TerraSAR-X radar images is presented. Both feature extraction and matching stages are still key issues when manipulating and processing remote sensing data for many applications like change detection and land use. Different blob, corner and scale based feature detectors have been tested during the keypoint extraction, descriptor construction and matching processes. A sub-sampling process is proposed in order to control the number of extracted keypoints from both optical and radar images for a real time processing with high resolution satellite images. After getting the pairwise matches between the input images, a full registration process is followed by applying image transformation (warping) and finally blending and compositing the registered version. A speckle noise removal techniques are performed on TerraSAR-X radar images as a pre-processing in order to reduce the number of false detections. The results showed a good performance for some of the detectors including SURF, Harris and FAST on optical and radar images. The tested optical and radar images are in different spatial resolutions and covering the area of Berlin Brandenburg airport.