SURE – The ifp Software for Dense Image Matching

Dense image matching methods enable the extraction of 3D surface geometry from images acquired from multiple views. Tpyical applications vary from aerial imaging, where such methods can be used to retrieve digital surface models with high density, up to cultural heritage data recording, where the acquisition of images using digital cameras represents an efficient method to retrieve 3D data for documentation purposes. For every application, the desired density and precision of the 3D surface information can be selected flexibly by choosing an appropriate image sensor and acquisition configuration. Within this paper, the dense image matching software SURE is presented, which has been developed by the Institute for Photogrammetry at the University of Stuttgart. It uses a multi-view stereo (MVS) approach, where first stereo pairs are matched against each other. This stereo matching step is based on the library libTSGM, which implements a modified version of the Semi Global Matching (SGM) algorithm – enabling the determination of 3D information for almost each pixel. The modification uses a hierarchical approach, which enables the processing of complex scenes with large depth variations with short processing time and low memory consumption. Within a second step, the results of image matching are fused by triangulating rays for multiple stereo models at once. This improves the precision of object points, but also enables the rejection of outliers as well as the determination of quality values for each 3D point. The whole implementation is parallelized and optimized for scalability. Thus, large datasets regarding image size and image count can be processed on common desktop PCs. SURE is available online for free for non-commercial use at http://www.ifp.uni-stuttgart.de/publications/software/.