Proceedings PaperThree-dimensional shape recovery by octree voting technique
|Format||Member Price||Non-Member Price|
The objective of this work is to develop a technique for 3D shape information recovery from multiple 2D images. Stereopsis is one of the most popular techniques for recovering 3D information. But it has an inherent and difficult problem with stereo matching, especially when the epipolar line is not parallel to the raster line. In this paper, we propose a new technique to avoid this problem by projecting back the feature points in images to the 3D space and by voting (just like for the Hough transform) with a well-contrived voting and evaluation rule. The octree representation of the object is adopted to enable multi resolutional analysis and to increase calculation efficiency. Experimental results are also described.