A COMPUTATIONAL MODEL OF DEPTH-PERCEPTION BASED HEADCENTRIC DISPARITY

It is now well established that depth is coded by local horizontal dis parity and global vertical disparity. We present a computational model which explains how depth is extracted from these two types of dispari ties. The model uses the two (one for each eye) headcentric directions of binocular targets, derived from retinal signals and oculomotor sig nals. Headcentric disparity is defined as the difference between headc entric directions of corresponding features in the left and right eye' s images. Using Helmholtz's coordinate systems we decompose headcentri c disparity into azimuthal and elevational disparity. Elevational disp arities of real objects are zero if the signals which contribute to he adcentric disparity do not contain any errors. Azimuthal headcentric d isparity is a ID quantity from which an exact equation relating distan ce and disparity can be derived. The equation is valid for all headcen tric directions and for all binocular fixation positions. Such an equa tion does not exist if disparity is expressed in retinal coordinates. Possible types of errors in oculomotor signals (six) produce global el evational disparity fields which are characterised by different gradie nts in the azimuthal and elevational directions. Computations show tha t the elevational disparity fields uniquely characterise both the type and size of the errors in oculomotor signals. Our model uses a measur e of the global elevational disparity field together with local azimut hal disparity to accurately derive headcentric distance throughout the visual field. The model explains existing data on whole-field dispari ty transformations as well as hitherto unexplained aspects of stereosc opic depth perception. (C) 1998 Elsevier Science Ltd. All rights reser ved.