GRASP Lab Seminar 2003-2004

February 6, 11:00 AM, Levine Hall 307, hosted by Jianbo Shi.

David Jacobs
University of Maryland

Recognizing and Reconstructing Objects When the Lighting Changes

Abstract: Variations in lighting can have a significant impact on the appearance of an object. In this talk I will discuss novel characterizations of this variability for the case of Lambertian(non-shiny) and specular (shiny) objects. First, we show that almost all the appearance of Lambertian objects is determined by the first nine components of the lighting when represented as spherical harmonics; these correspond to the diffuse components of lighting. Given a 3D model, this allows us to analytically derive a 9D linear subspace that can be used to build efficient object recognition algorithms. We apply these ideas to face recognition, and describe experiments on a data base of 42 3-D models of faces, with 300 query images. We also show how they can be used to reconstruct the shape of Lambertian objects using images taken under changing, uncontrolled lighting conditions. Next, to handle non-Lambertian reflectance we introduce a simple model of specular reflection that allows us to account for highlights produced by a dominant, compact light source. We use this model to identify objects such as wine glasses, whose appearance is characterized by little more than highlights. Combining these methods we can identify objects made from a wide range of materials such as ceramics, that mix Lambertian and non-Lambertian reflectance. Joint work with: Ronen Basri, The Weizmann Institute Margarita Osadchy, NEC Labs Ravi Ramamoorthi, Columbia University

Biography: David W. Jacobs received the B.A. degree from Yale University in 1982. From 1982 to 1985 he worked for Control Data Corporation on the development of data base management systems, and attended graduate school in computer science at New York University. From 1985 to 1992 he attended M.I.T., where he received M.S. and Ph.D. degrees in computer science. From 1992 to 2002 he was a Research Scientist and then a Senior Research Scientist at the NEC Research Institute. In 1998 he spent a sabbatical at the Royal Institute ofTechnology (KTH) in Stockholm. Since 2002, he has been an Associate Professor of computer science at the University of Maryland, College Park. Dr. Jacobs' research has focused on human and computer vision, especially in the areas of object recognition and perceptual organization. He has also published articles in the areas of motion understanding, memory and learning, and computational geometry. He and his co-authors received honorable mention for the best paper award at CVPR 2000.

full schedule