A LBP Based Correspondence Identification Scheme for Multi-view Sensing Network
This work addresses a domain-specific problem in computer vision for applications like 3D reconstruction and image search, but it appears incremental as it builds on existing techniques like normalized cross correlation and loopy belief propagation.
The paper tackles the problem of identifying correspondences between two views of regular RGB cameras in real-time by first using normalized cross correlation for sparse matching and then applying loopy belief propagation for dense correspondence identification. The results demonstrate superb accuracy and precision that outperform state-of-the-art methods in computer vision.
In this paper, we describes a correspondence identification method between two-views of regular RGB camera that can be run in real-time. The basic idea is first applying normalized cross correlation to retrieve a sparse set of matching pairs from image pair. Then loopy belief propagation scheme is applied to the the set of possible candidates to densely identify correspondences from different views. The experiment results demonstrate superb accuracy and precision that outperform the state-of-the-art in the computer vision field. Meanwhile, the implementation is simple enough that can be optimized for real-time performance. We have given the detailed comparison of existing approaches and show that this method can enable various practical applications from 3D reconstruction to image search.