ACM Computing Surveys (CSUR)
Probabilistic regularisation and symmetry in binocular dynamic programming stereo
Pattern Recognition Letters - In memory of Professor E.S. Gelsema
Visual Correspondence Using Energy Minimization and Mutual Information
ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
Thrift: Local 3D Structure Recognition
DICTA '07 Proceedings of the 9th Biennial Conference of the Australian Pattern Recognition Society on Digital Image Computing Techniques and Applications
Stereo Processing by Semiglobal Matching and Mutual Information
IEEE Transactions on Pattern Analysis and Machine Intelligence
Keypoint Detection and Local Feature Matching for Textured 3D Face Recognition
International Journal of Computer Vision
Fast geometric point labeling using conditional random fields
IROS'09 Proceedings of the 2009 IEEE/RSJ international conference on Intelligent robots and systems
International Journal of Computer Vision
Are we ready for autonomous driving? The KITTI vision benchmark suite
CVPR '12 Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Learning to efficiently detect repeatable interest points in depth data
ECCV'12 Proceedings of the 12th European conference on Computer Vision - Volume Part I
Hi-index | 0.00 |
The use of Points-of-Interest can significantly reduce computational complexity and memory required for high-level vision tasks. We describe a simple and effective technique to detect Points-of-Interest from noisy disparity maps generated by a real-time stereo system, which considers all sources of information, i.e. the disparity map and the left and right images of a stereo pair. The first step in our approach assigns states to each point and marks possible border points. Then we detect left and right image edges. Matching these three sources of information leads to the Points-of-Interest, which we call Triple Edge points. The final step identifies significant points by projecting the detected points to real-world 3D space. Generation of the Triple Edge points is computationally more efficient than other techniques of Point-of-Interest detection. Some simple experimental results show that the Triple Edge points are a reliable source of information for identifying multiple objects in low-texture scenes with noisy disparity maps. We argue that they can be used in other high-level tasks such as object extraction, pose estimation, recognition and 3D reconstruction.