- Photogrammetric Surface Reconstruction from Imagery
Authors: Mathias Rothermel & Konrad Wenzel
SURE is a software solution for multi-view stereo, which enables the derivation of dense point clouds from a given set of images and its orientations. Up to one 3D point per pixel is extracted, which enables a high resolution of details. It is based on libtsgm, which implements the core functionality for image rectification, dense matching and multi-view triangulation, and provides a C/C++ API.
SURE can handle image collected by various types of sensors and can be utilized for close range, UAV and large frame aerial datasets. It scales well to datasets with large image numbers and large image resolution (e.g. >200MP aerial imagery). Also imagery with high dynamic range (e.g. 16 Bit depth) can be used. Initial depth information of the scene is not required, which enables the use within fixed calibrated camera setups. The efficiency of processing is increased by utilizing parallel processing and hierarchical optimization.
|St. Andrews, Scotland. 43 images with 5 Megapixels resolution. Result from SURE: 43 Mio 3D points.
|Hessigheim, Germany. 3D point cloud derived from a medium format aerial camera (IGI DigiCam, 50 Megapixels)|
The input of SURE is a set of images and the corresponding interior and exterior orientations. This orientation can be derived either automatically (e.g. by Structure from Motion methods) or by using classical image orientation approaches. So far interfaces to common image orientation software, such as Bundler (.out), VisualSFM (.nvm), Trimble/INPHO Match-At (aerial imagery, .prj), are provided. The output of the software is point clouds or depth images.
SURE is a multi-stereo solution. Therefore, single stereo models (using 2 images) are processed and subsequently fused. Within a first step images are undistorted and pair-wise rectified. Within a second step, suitable image pairs are selected and matched using a dense stereo method similar to the Semi Global Matching (SGM) algorithm (Rothermel et. al., 2013 ). In contrast to the original SGM method as published in (Hirschmüller, 2008 ) a more time and memory efficient solution is implemented. Within a triangulation step the disparity information from multiple images is fused and 3D points or depth images are computed. By exploiting the redundancy of multiple disparity estimations precision of single depth measurements is increased and blunders are efficiently filtered. For detailed information refer to our publication .
We are grateful for any critics, suggestions, remarks and bug reports! If you have generated some nice results, we would appreciate if you could send us screenshots, in order to add them to our gallery! Happy matching!
For more images see our gallery!
Terms and Conditions
The free version of SURE is provided for research and non-commercial use only. In case you use this software in your research, please cite reference . Within the free version, the number of images is limited to 1000, while the sum of the resolution for all images is limited to 5000 Megapixels in total. Furthermore, the estimated precision per point is only available in the full version. If you would like to use the software commercially and without limitations, please get in touch with us using the email address below. See the readme file provided with the software for further details concerning the license.
Contact & Licenses
For commercial licenses and more information please visit our website nframes.com.
You can download SURE for non-commercial use using the links below:
- Hello World!
Getting started & manual
- Extract archive to folder
- Ready! If you want to use it from your command line, it might be useful to add the directory to you PATH environment variable, so that you can call the executable in every directory.
See installation.txt provided with the linux package.
SURE currently depends on the following third party libraries:
- OpenCV version 2.4.3 ( http://opencv.willowgarage.com)
- Boost ( http://www.boost.org)
- LibLas ( http://www.liblas.org)
- libdaisy version1.7.1 ( http://cvlab.epfl.ch/software/daisy)
Image formats: jpg, png, tif (also tiled and 16 Bit depth)
Orientation formats: Ori (documentation below), Bundler (.out), VisualSFM (.nvm), Trimble/Inpho (.prj)
- point clouds: .las format (ASPRS point cloud format); ascii xyz
- range images
We wrote a paper  about image acquisition for multi-view stereo surface
reconstruction, which you can find here:
SURE was used in multiple projects and studies. Results for UAV and large frame aerial image collections were published in  and . A second implementation of a hierarchical SGM approach was used for large scale cultural heritage projects with high resolution requirements, as published in .
 Rothermel, M., Wenzel, K., Fritsch, D., Haala, N. (2012). SURE: Photogrammetric Surface Reconstruction from Imagery. Proceedings LC3D Workshop, Berlin ,December 2012
 Hirschmüller, H., 2008. Stereo Processing by Semiglobal Matching and Mutual Information. IEEE Transactions on Pattern Analysis and Machine Intelligence 30, 328-341.
 Wenzel, K., Rothermel, M., Fritsch, D., and Haala, N.: Image Acquisition and Model Selection for Multi-View Stereo, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XL-5/W1, 251-258, doi:10.5194/isprsarchives-XL-5-W1-251-2013, 2013.
 Rothermel, M., Haala, N., 2011. Potential of Dense Matching for the Generation of High Quality Digital Elevation Models. In ISPRS Proceedings XXXVII 4-W19
 Haala, N. & Rothermel, M., 2012. Dense Multi-Stereo Matching for High Quality Digital Elevation Models, in PFG Photogrammetrie, Fernerkundung, Geoinformation. Jahrgang 2012 Heft 4 (2012), p. 331-343.
 Wenzel, K., Abdel-Wahab, M., Cefalu, A., and Fritsch, D.: High-Resolution Surface Reconstruction from Imagery for Close Range Cultural Heritage Applications, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XXXIX-B5, 133-138, doi:10.5194/isprsarchives-XXXIX-B5-133-2012, 2012.