The Application of Approximation and Complexity Theory Methods to the Solution of Computer Vision Problems

Hatzitheodorou, Michael

We survey aspects of approximation and complexity theory and their application to the numerous computer vision problems that require an approximate solution because only partial information is available. We consider ill-posed computer vision problems and the methods that can be employed towards reformulating them as well-posed. We are particularly interested in the surface reconstruction problem that is encountered in the construction of the 2 1/2-D sketch, and which has been formulated and solved using different methods. We apply regularization theory, information-based complexity, and other methods to the solution of this problem. Finally, the shape from shadows problem is formulated and the optimal error algorithm is constructed and analyzed.



More About This Work

Academic Units
Computer Science
Department of Computer Science, Columbia University
Columbia University Computer Science Technical Reports, CUCS-335-88
Published Here
December 9, 2011