Variation among human observers in detecting visual texture differences: is the length of the boundary between different textured regions or the areas covered by them more important?

Abstract

Textures defined by monochromatic texels were used to measure the ability of human observers to distinguish a ‘primary’ texture covering one area from another texture, the ‘background’ texture, in a neighboring or embedding area when both textures are presented briefly and simultaneously. A mask containing random gray-scale dots is presented after the stimulus to help prevent after-imaging. The texels used were carefully designed so that the discrimination could not be made using relative brightness cues only. The textures were selected so that the primary texel elements were either spread uniformly over a limited region, or were systematically interspersed with texels of the background texture. Texture combinations were also used that varied the texel density gradient at the boundary between the figure and background textures. We extend our earlier result which showed significant variation among the human observers by showing that the variation cannot be explained by postulating that the ‘mechanisms’ responsible for texture detection operate at different speeds for different observers. We also present results that show the relative importance of the size of the area covered by the figure, and the ‘effective’ length of the enclosing boundary of the primary texture. The results are interpreted in terms of some published models that are in vogue for determining texture boundaries.