Abstract
The aim of this study was to examine the contribution of stereo information to recognition performance of multi-part objects across changes in stimulus viewpoint. A recognition task was performed under either monocular or stereo viewing conditions where participants first learned a set of 12 novel 3D objects from three different viewing angles (Learning Phase) and then discriminated previously learned objects from visually similar distractors (Test Phase). The previously learned objects were presented from either the same viewpoints AS shown in the learning phase or different viewpoints falling between the trained views (Interpolated) or outside of the trained views (Extrapolated). Behavioural measures were obtained from the recognition task to compare performance differences between the monocular and stereo viewing conditions and viewpoint rotations. The accuracy scores revealed an advantage for correctly discriminating previously learned objects viewed from unfamiliar viewing angles when stereo information was available. The results also showed an advantage for recognising interpolated over extrapolated views and the presence of a significant interaction between viewing condition and viewpoint. This suggests that stereo information can modulate viewpoint costs possibly via the integration of stereo-defined depth cues during view interpolation.
Meeting abstract presented at VSS 2012