Information Technology Reference
In-Depth Information
7
Conclusions and Future Work
We have summarised several techniques for effective stereo correspondence from
2-D images. These systems can reasonably deal with the common problems such as
wide-baseline, clutters, and light changing. In the maximum likelihood estimation
approach, flat surfaces are extracted from the scenes by analysing the video content,
e.g. correspondence and 3-D recovery. An iterative RANSAC plane fitting scheme
was also presented. In the segment based stereo algorithm, the image patches look
neat in most cases. This is due to the colour segmentation before the disparity maps
are estimated. In the connectivity based scheme, the stereo correspondence is inte-
grated with shape segmentation. The shape segmentation is used to enhance the per-
formance of estimating the disparity maps. The last one is the wide-baseline stereo
strategy incorporating local descriptors. This algorithm can effectively handle some
significant wide-baseline cases. In spite of their success, these systems also reveal
their weakness in certain circumstances. One of the disadvantages is that, in many
cases, the local noise still evidently appears and somehow affects the structure rep-
resentation of the scene or objects. This weakness may be tackled if prior knowledge
of these details can be used after necessary training of a local classifier.
Acknowledgements.
This work was supported by European Commission under Grant FP6-
045189-STREP (RUSHES).
References
1. Sivic, J., Zisserman, A.: Video google: a text retrieval approach to obejct matching in
videos. In: Proc. of Ninth IEEE International Conference on Computer Vision, pp. 1470-
1477 (2003)
2. Davis, M.: An iconic visual language for video annotation. In: Proc. of IEEE Symposium
on Visual Language, pp. 196-202 (1993)
3. Rushes project
deliverable
D5,
requirement
analysis and
use-cases
definition for
professional content creators or providers and home-users (2007),
4. Tangelder, J., Veltkamp, R.: A survey of content based 3d shape retrieval methods. In:
Proc. of International Conference on Shape Modeling, pp. 145-156 (2004)
5. Ohbuchi, R., Kobayashi, J.: Unsupervised learning from a corpus for shape-based 3d
model retrieval. In: Proc. of the 8th ACM international workshop on Multimedia infor-
mation retrieval, New York, NY, USA, pp. 163-172 (2006)
6. Bartoli, A., Sturm, P.: Structure-from-motion using lines: Representation, triangulation,
and bundle adjustment. Computer Vision and Image Understanding 100(3), 416-441
(2005)
7. Zhou, H., Wallace, A., Green, P.: A multistage filtering technique to detect hazards on
the ground plane. Pattern Recognition Letters 24(9-10), 1453-1461 (2003)
8. Klaus, A., Sormann, M., Karner, K.: Segment-based stereo matching using belief propa-
gation and a self-adapting dissimilarity measure. In: Proc. of 18th International Confer-
ence on Pattern Recognition, pp. 15-18 (2006)
