Authors: [tex2html_wrap4110]K. Wu, M.D. Levine
Investigator username: levine
Subcategory: computer vision
Object data obtained from multiple viewpoints contain much more shape information of a complicated object than data taken from a single view. Various techniques for multiview integration have been presented for many years. However, work on using the multiview data to build high-level descriptions have rarely appeared in the literature. In this work, we are exploring the decomposition of a 3D object into human notions of parts using multiview-integrated data. The criterion used for decomposition is the sharp discontinuity on object surfaces. The complex problem of handling 3D data is converted to a simpler task dealing with a stack of 2D slices. The various constraints on 3D parts manifested in a single 2D slice and among these slices are being studied.