SURFACE CONSTRAINT: AN APPROACH TO SOLVING THE STEREO DEPTH PROBLEM ((J. Geier)) Department of General Psychology, Lorand Eotvos University, Budapest, Hungary.

Purpose A global computational model and computer simulation for solving the stereo correspondence problem is presented. A new constraint has been developed to avoid false targets.

Methods The method is based on the following definition of surface constraint: with a given stereo image pair and known camera-parameters, if the left and right image are projected to the original 3D object in its original position, then the same texture will definitely appear on the 3D surface - except in the occluded regions. The computational goal: with a presented stereo image pair, to find the representation of the 3D surface that provides the best satisfaction of the surface constraint. The computational algorithm is an improved version of a method presented last year (J. Geier, 1995, ARVO95 #1722). This algorithm includes the surface constraint in order to test the goodness of the matching. Further constraints (i.e. smoothness, uniqueness, ordering etc.) have not been employed.

Results The computer algorithm has been tested on real life images and computer generated random dot stereograms. The matching is correct at nearly all 3D points.

Conclusions The model gives a good explanation for human stereo vision. The surface constraint is sufficient in itself, no other constraints are necessary to avoid false targets.

Supported by MAKA J.F. no. 360