Local Grayvalue Invariants for Image Retrieval
IEEE Transactions on Pattern Analysis and Machine Intelligence
On accurate and robust estimation of fundamental matrix
Computer Vision and Image Understanding
Pattern recognition in Practice VI
Pattern Recognition Letters - Special issue on pattern recognition in practice VI
Pattern Recognition in Practice IV: Multiple Paradigms, Comparative Studies and Hybrid Systems: Proceedings of an International Workshop Held in Vlieland, the Netherlands, 1-3 June 1994
Triangle: Engineering a 2D Quality Mesh Generator and Delaunay Triangulator
FCRC '96/WACG '96 Selected papers from the Workshop on Applied Computational Geormetry, Towards Geometric Engineering
Making Good Features Track Better
CVPR '98 Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
Closed-form connectivity-preserving solutions for motion compensation using 2-D meshes
IEEE Transactions on Image Processing
Hierarchical 2-D mesh representation, tracking, and compression for object-based video
IEEE Transactions on Circuits and Systems for Video Technology
Point Matching under Large Image Deformations and Illumination Changes
IEEE Transactions on Pattern Analysis and Machine Intelligence
Target tracking algorithm based on optical flow method using corner detection
Multimedia Tools and Applications
Hi-index | 0.00 |
In applications such as video mosaicing, foreground/background separation and camera pose estimation, feature correspondences are the fundamental building blocks upon which the method operates. The success or failure of these methods depends strongly on the quality of the feature correspondences obtained from the video sequence. We present a method that attempts to improve the quality of feature matches in cases when the video sequence is rich in perspective effects and 3-d camera induced motion from static scenes. This is achieved by fitting an active mesh to the sequence and then matching not features but mesh induced planar patches. The optic flow of each mesh element votes for the overall motion of its corner points to achieve a high quality fit of the mesh to each subsequent frame. We find that we obtain feature correspondences that allow us to fit 3-d metrics to a higher degree of accuracy than that obtained when using a standard feature tracker.