Region-based Segmentation on Evolving Surfaces

Anthony J. Yezzi
ayezzi@ece.gatech.edu
Georgia Institute of Technology

Project summary

Many real objects (especially man-made) are made by composing different materials, and therefore possess piecewise constant reflectance properties. Consequently, the object radiances are piecewise smooth under general illumination conditions. We are interested in developing algorithms to estimate the shape and radiance from collections of images for such scenes. The approach we take is in the general framework of comparing images with model. (See related work on Stereoscopic Segmentation and Multi-view Stereo Beyond Lambert.) We model the scene radiance with piecewise smooth functions whose discontinuities are curves on the unknown surface. The estimation process thus involves performing region-based segmentation on evolving surfaces.

Related publications

H. Jin, A. Yezzi and S. Soatto.
Region-based Segmentation on Evolving Surfaces with Application to 3D Reconstruction of Shape and Piecewise Smooth Radiance.
Technical Report, CSD-TR04-0004, University of California at Los Angeles, 2004.
In preparation for submission to Intl. J. Computer Vision.
H. Jin, A. Yezzi and S. Soatto.
Region-based Segmentation on Evolving Surfaces with Application to 3D Reconstruction of Shape and Piecewise Constant Radiance.
To appear in Euro. Conf. on Computer Vision, May 2004.
A. Yezzi, S. Soatto, H. Jin, A. Tsai, and A. Willsky.
The Mumford-Shah functional from Segmentation to Stereo.
In Geometric Level Set Methods in Imaging, Vision and Graphics, S. Osher and N. Paragios (Eds), Springer Verlag, 2003.
H. Jin.
Variational Methods for Shape Reconstruction in Computer Vision.
Ph.D. thesis, Electrical Engineering Department, Washington University, August 2003.
H. Jin, A. Yezzi, Y. Tsai, L. Cheng and S. Soatto.
Estimation of 3D Surface Shape and Smooth Radiance from 2D Images: a level set approach.
Journal of Scientific Computing, 19(1-3), pages 267-292, December 2003.

Sponsors

This project is supported by NSF grants IIS-0208197 and CCR-0133736, ONR grant N00014-02-1-0720, AFOSR grant F49620-03-1-0095.

Last updated on February 4, 2004.
BACK to Hailin Jin's homepage