Information Technology Reference
In-Depth Information
Chapter 13
The Use of Color Information in Stereo Vision
Processing
Wided Miled and Béatrice Pesquet-Popescu
Abstract.
Binocular stereovision is the process of recovering the depth informa-
tion of a visual scene, which makes it attractive for many applications like 3-D
reconstruction, multiview video coding, safe navigation, 3-D television and free-
viewpoint applications. The stereo correspondence problem, which is to identify the
corresponding points in two or more images of the same scene, is the most impor-
tant and difficult issue of stereo vision. In the literature, most of the stereo match-
ing methods have been limited to gray level images. One information that has been
largely neglected in computational stereo algorithms, although typically available in
the stereo images, is color information. Color provides much more distinguishable
information than intensity values and can therefore be used to significantly reduce
the ambiguity between potential matches, while increasing the accuracy of the re-
sulting matches. This would largely profit to stereo and multiview video coding,
since efficient coding schemes exploit the cross-view redundancies based on a dis-
parity estimation/compensation process. This chapter investigates the role of color
information in solving the stereo correspondence problem. We test and compare
different color spaces in order to evaluate their efficiency and suitability for stereo
matching.
1
Introduction
Research on multiview technologies has been widely enhanced recently, covering
the whole media processing chain, from capture to display. A multiview video sys-
tem consists in generating multiple views by capturing from different viewpoints the
same scene via a set of multiple cameras. By presenting the corresponding image
