Multi-view stereo beyond lambert

  • Authors:

  • Hailin Jin;Stefano Soatto;Anthony J. Yezzr

  • Affiliations:

  • Electrical Engineering Department, Washington University, St. Louis, MO;Computer Science Department, University of California, Los Angeles, CA;Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA

  • Venue:
  • CVPR'03 Proceedings of the 2003 IEEE computer society conference on Computer vision and pattern recognition
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

We consider the problem of estimating the shape and radiance of an object from a calibrated set of views under the assumption that the reflectance of the object is non-Lambertian. Unlike traditional stereo, we do not solve the correspondence problem by comparing image-to-image. Instead, we exploit a rank constraint on the radiance tensor field of the surface in space, and use it to define a discrepancy measure between each image and the underlying model. Our approach automatically returns an estimate of the radiance of the scene, along with its shape, represented by a dense surface. The former can be used to generate novel views that capture the non-Lambertian appearance of the scene.