DETECTION OF ORIENTATIONALLY MULTIMODAL TEXTURES

Oriented textures were produced with the use of probability density fu nctions modulated sinusoidally over orientation. Orientational contras t sensitivity functions (OCSFs) for a task involving the discriminatio n of these patterns from orientationally-random textures were found fo r several human observers. An inverse Fourier transform of this OCSF y ielded a weighting function, or filter, defined over orientation. The weighting function is broad, with a half-height full-width of 34 deg. This orientational filter was able to predict human performance in fur ther discrimination tasks employing a variety of probability density f unctions over orientation.