A Locally Global Approach to Stereo Correspondence
Abstract
A novel approach to deal with the stereo correspondence \nproblem induced by the implicit assumptions made \nby cost aggregation (CA) strategies is proposed. CA relies \non the implicit assumption that disparity varies smoothly \nwithin neighboring points except at depth discontinuities \nand state-of-the-art CA strategies adapt their support to image \ncontent by classifying each pixel based on geometric \nand photometric constraints. Our proposal explicitly models \nthis behavior from a different perspective, by gathering \nfor each point, multiple assumptions that locally would be \nmade by a hypothetical variable CA strategy. This framework \nenables to derive a function that locally captures the \nplausibility of the underlying geometric and photometric \nconstraints independently enforced by supports of neighboring \npoints. Experimental results confirm the effectiveness \nof our proposal.
Cite
Text
Mattoccia. "A Locally Global Approach to Stereo Correspondence." IEEE/CVF International Conference on Computer Vision Workshops, 2009. doi:10.1109/ICCVW.2009.5457496Markdown
[Mattoccia. "A Locally Global Approach to Stereo Correspondence." IEEE/CVF International Conference on Computer Vision Workshops, 2009.](https://mlanthology.org/iccvw/2009/mattoccia2009iccvw-locally/) doi:10.1109/ICCVW.2009.5457496BibTeX
@inproceedings{mattoccia2009iccvw-locally,
title = {{A Locally Global Approach to Stereo Correspondence}},
author = {Mattoccia, Stefano},
booktitle = {IEEE/CVF International Conference on Computer Vision Workshops},
year = {2009},
pages = {1763-1770},
doi = {10.1109/ICCVW.2009.5457496},
url = {https://mlanthology.org/iccvw/2009/mattoccia2009iccvw-locally/}
}