High frequency textures provide better support for shape-from-shading than low frequency textures

Observers perceive a sinusoidally shaded texture as a corrugated surface even when the texture elements themselves undergo no geometric distortions (Schofield, Heese, Rock & Georgeson, 2006, Vision Research, 46, 3462–3482). Using a similar two-point probe task but Gabor noise textures, we varied the dominant spatial frequency of the texture (from 1.5 to 12 c/deg) and found that high frequency textures support a more robust percept of shape-from-shading than do low frequency textures. Given that our sinusoidal shading patterns were themselves low frequency (0.5 c/deg) we were concerned that this difference may be due to masking. That is, the low frequency textures might simply have reduced the visibility of the shading patterns. To control for this we varied the dominant orientation of the textures so as to reduce their ability to mask the shading pattern; this had no affect. Reducing the spatial-frequency bandwidth of the textures, which should reduced masking, also had no affect. Multiplicative shading of an albedo textured surface produces a change in local mean luminance coupled with a change local luminance amplitude (AM). Schofield et al. (2006) showed that this AM cue modulates the perception of shape-from-shading. Given that AM is a second-order cue requiring comparisons across pairs of pixels, our results are consitent with the idea that second-order processes receive most of their input from high-frequency channels (Dakin & Mareschal, 2000, Vision Research, 40, 311–329). We speculate that when the carrier texture is high frequency, AM is detected well and thus supports shape-from-shading. When the carrier is low frequency AM is detected less well and consequently shape-from-shading is inhibited.