Abstract:

This paper presents a computational model for stereopsis. Laplacian of Gaussian filters are used to extract spatial features and a set of one-octave Gabor filters is used to extract orientation information from a scene. A mapping from a 3D space onto two 2D image planes is derived using a Gaussian sphere model. The determinant of the Jacobian of the mapping is derived and matching is performed using zero-crossings associated with their orientation information. Relaxation labelling is used as a co-operative process for solving the correspondence problem.

Document Type: Research article

DOI: 10.1163/156856896X00024

Affiliations: 1: School of Computer Science and Engineering, University of New South Wales, Sydney 2052, Australia 2: AI Lab. Computer and Information Science, University of Otago, Dunedin, New Zealand