Segmented Shape Descriptions from 3-View Stereo

Abstract: We address the recovery of segmented, 3-0 descriptions of an object from intensity images. We use three views of an objectfrom slightly different viewpointsas our input. For each image we extract a hierarchy of groups based on proximity, parallelism and symmetry in a robust manner. Thegroups in the three images are matched by computing the epipolar geometry. For each set of matched groupsfrom the three images, we then label the contours of the groups as “true”or “limb” edges. Usingthe informationaboutgroups, the label associated with their contours and projective properties of subclasses of Generalized Cylinders,we infer the 3D structure of these groups. Theproposed methodnot onlyallows robust shape recovery but also produces segmented parts. Our approach can also deal with groups generated as a result of textureor shadowson the object. Wepresent results on real images of moderately complex objects. Key Words:Shape description, Grouping,Stereo, Edge Labeling, Generali...

• Type: Conference
• Year: 1995
• Where: ICCV
• Authors: Parag Havaldar, Gérard G. Medioni
• Keywords: Computer Vision, ICCV 1995, Images, Intensity Images, Robust Manner