Phase-Based Disparity Estimation in Binocular Tracking

Atsuto Maki, Tomas Uhlin and Jan-Olof Eklundh

Proc. 8th Scandinavian Conference on Image Analysis (K. Heia K. A. Høgdra, B. Braathen, ed.), (Tromsø, Norway), pp. 1145--1152, Norwegian Society for Image Processing and Pattern Recognition, May 1993.

Abstract

Binocular robots whose cameras can be independently directed require some mechanism for verging; i.e. aiming both cameras at a fixation point. Since the desired vergence angle is directly related to target distance, some sensory cue to give depth information is required to the vergence system. The estimationnadabinoculaCVAP/r disparity is a fundamental precursor to the depth estimation. Algorithms using phase provide a new promising method for example based on the output of Gabor filters for disparity estimation. In this article, at first, we give a brief description of the phase-based algorithm, including an extension to handle the "wrap-around" problem, and a coarse-to-fine strategy "the hierarchical method with pixel-shift" is presented. For binocular tracking, computational time is restrict. Nevertheless, generally heavy calculations are required to get an output through Gabor filters. For that reason, a step-wise fast filter is examined and it is shown to be a substitute for the Gabor filter. Finally the algorithm with the fast filter is applied to a tracking control on the KTH head-eye system.

Full paper: PostScript 0.9Mb


Atsuto Maki <maki@bion.kth.se>
Tomas Uhlin <tomas@bion.kth.se>
Jan-Olof Eklundh <joe@bion.kth.se>