O'Brien, Justin M. D.; (1998) Texture and motion cues in slant perception. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Texture_and_motion_cues_in_sla.pdf Download (24MB) | Preview

Abstract

A series of experiments was conducted to determine the relative weights of motion and texture cues in the perception of surface slant, together with the properties which effect this weighting. The stimuli used were ray-traced image sequences of surfaces oriented in depth, in which texture and motion properties could be varied independently. It has previously been established that motion parallax or shearing (Rogers and Graham 1979) and motion about an axis normal to the line of sight (Young, Landy and Maloney 1993) both produce strong impressions of depth. Since texture gradients and motion parallax provide the same information about surface slant, it was this latter type of motion that was used throughout the experimental programme. When texture and motion cues were consistent, depicting the same surface slant, it was determined that the addition of a 2D shearing and compressive motion cue increased perceived slant by only 3°, leaving slant discrimination thresholds unchanged. The most significant determinant of perceived slant was the nature of the texture cue. A ID texture pattern gave higher thresholds and lower perceived slant angle than a 2D texture pattern. Further studies on the properties of texture determined that texture spatial frequency had little effect on perceived slant for a given field of view, as did using a combination of frequencies to create a more complex texture pattern. It was shown that slant discrimination threshold was a function of surface slant, with lower thresholds the higher the slant. It was found, furthermore, that while ID horizontal textures generally gave higher thresholds and lower perceived slant angle than ID vertical textures, this disparity was itself a function of slant angle, with higher disparities at smaller slant angles. When the slant depicted by texture and that by motion were slightly different, texture was the more highly weighted cue. When texture slant was constant and motion slant varied by up to 12°, perceived slant varied by 8% of the change in motion slant, while when motion slant was constant and texture slant varied, perceived slant varied by 85% of the change in texture slant. The regularity of the texture pattern had no significant effect on this weighting. In a further study, it was demonstrated that thresholds for perceiving changes in ID and 2D velocity gradients were only 30% higher than thresholds for changes in ID and 2D spatial frequency gradients. It is concluded that when texture and motion cues provide the same information about surface slant, it is the texture cue which dominates in slant perception. This contrasts with results found when the motion cue consists of a rotation about and axis normal to the line of sight (Young et al. 1993), where texture and motion cues can be found to have approximately equal weights.

Download activity - last month

Export as XMLJSONCSV

Download activity - last 12 months

Export as XMLJSONCSV

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item