This paper presents a detailed description of a real-time correlation-based stereo algorithm running completely on the graphics processing unit (GPU). This is important since it allows to free up the main processor for other tasks including high-level interpretation of the stereo results. We first introduce a two-view stereo algorithm that includes some advanced features such as adaptive windows and cross-checking. Then we extend it using a plane-sweep approach to allow multiple frames without rectification. By taking advantage of advanced features of recent GPUs the proposed algorithm runs in real-time. Our implementation running on an ATI Radeon 9800 graphics card achieves up to 289 million disparity evaluations per second including all the overhead to download images and read-back the disparity map, which is several times faster than commercially available CPU-based implementations.
|Number of pages||12|
|State||Published - Feb 2005|
Bibliographical noteFunding Information:
The authors would like to thank ATI and NVIDIA for their generous hardware donations and technical support. The work is supported in part by NSF grant IIS-0313047, IIS-0448185, and the office of research at the University of Kentucky.
ASJC Scopus subject areas
- Signal Processing
- Computer Vision and Pattern Recognition
- Electrical and Electronic Engineering