アイテム詳細

要旨

Prior to analyses using transformational apparent motion it was thought that form cues play little if any role in determining the perceived direction of apparent motion. Transformational apparent motion established, however, that there is not only perceptual 'structure-from-motion' but also 'motion-from-structure.' In particular, certain scenes must be segmented into overlapping objects before motion can be perceived to have taken place over those objects. Thus there is a parsing problem that logically precedes the traditional correspondence-matching problem. In the past I have shown how this segmentation takes place over flat figures that transform their shape. Here I show that segmentation takes place over volumes. In particular, when a 3-D form has been specified in the first frame of a two-frame apparent-motion display, the motion that is perceived upon presentation of the second frame is consistent with the 3-D form specified in the first frame, even when the form appearing in the second frame can be seen to have multiple 3-D interpretations. Thus parsing does not take place in each image independently of how previous images were segmented. Parsing must be understood as a 3-D spatiotemporal process that subsumes both the spatial parsing problem and the temporal correspondence-matching problem.