The orientational resolution of human texture perception.

A major determinant of human texture segregation and discrimination is the orientational content of the stimuli used. We have investigated the ability of observers to resolve features defined in the orientation domain in a variety of textures. It was found that features had to be separated by at least 13 deg for subjects to discriminate orientationally bimodal textures from same-variance unimodal textures. For larger separations, the determinant of performance was the magnitude of the central "dip" in the probability density functions determining the bimodal textures. Resolution performance can be modelled by assuming that a filtering process over orientation demodulates the central dip in the bimodal texture and that discrimination depends on criterion depth in the resulting function. Such modelling produces relatively broad estimates of the bandwidth ranging from about 10-20 deg. Performance was similar for both line and Gabor micropattern stimuli.