| Abstract: | The modern "textbook" view of visual perception contains an inherent paradox. On the one hand, it claims that relatively simple edge-extraction processes requires a stimulus exposure of approximately 50 ms. On the other hand, it says that the identification of objects in photographs and line-drawings can be highly accurate with exposure durations as short as 100 ms. It is tempting to conclude that all the difficult work of perception occurs in the 50 ms that elapse between when these two tasks are accomplished. This article argues against this view, suggesting instead that much more than edge-extraction is accomplished by the early visual processes. To illustrate this view, a computational model is described that is capable of recovering the 3-D orientation of objects from some line-drawings, rapidly and in parallel. Data from recent visual search experiments with human observers are presented in support of this model and the implications of this view for the "textbook" view are discussed. |
