Abstract

This study investigated the ability of humans to combine information about the position and orientation of straight edges from two or more regions of their spatial frequency spectra. The stimuli were luminance edges and straight lines decomposed into spatial frequency bands by using Laplacian pyramids. Different combinations of the frequency components contaminated by additive noise were presented at either the same or different spatial locations. The task was to judge the positions and orientations of the edges with respect to a probe point. Depending on the condition, the optimal fusion strategy was either a linear or nonlinear additive weighted combination. In the nonlinear condition the weights at each location ought to depend on the information at that location. Although observers showed some sensitivity to the noise level in different bands, they performed suboptimally. Unlike in the case of motion perception, high frequencies in edges captured low frequency information.

© 1992 Optical Society of America