Designing visually rich, nearly random textures.

Camouflaging textures containing as in real life edges at all orientations, were designed by computer, then manually, for use in stereoscopic vision studies. In the manual procedure, the starting point is either a set of photographs (for instance, of barks) or a manually produced first-generation texture. Then patches are cut zigzagging and assembled into successive generations of textures. The absence of extended edges--straight or curved--and the local heterogeneity of the texture are important camouflaging factors, allowing curved surfaces to be covered with these textures without visible join. Small areas of a texture often suggest a scene, but when the areas are assembled, the suggestive power is lost, and the statistical properties of the texture then dominate. However, when symmetry is introduced (as in the Rorschach test), meaningful scenes emerge again.