Seminar Series

Snapshots of completion

One of the most important goals of vision is to recognize objects so that we can interact appropriately with them. Our everyday experience suggests that the visual system achieves this goal instantly and effortlessly: We seem to have no difficulty organizing complex scenes into meaningful units --objects -- and recognizing those objects despite dramatic changes in the visual stimulation reaching our eyes. However, the results of many experiments have shown that our phenomenology greatly underestimates the complexity of the processes subserving perceptual organization and object recognition. These lectures will focus on important component of perceptual organization and recognition, namely the perception of partly-occluded objects. Occlusion presents significant problems for organization and recognition because significant pieces of an object often are hidden from view, and the visible parts often are distributed across non-adjacent parts of the visual array. We will present evidence that completion of partly-occluded objects takes time, that the completion process is affected by spatio-temporal context, and that completed contours are used to compute global properties like shape. These results suggest that completion acts as a grouping mechanism enabling observers to use the relevant parts of the stimulus more efficiently for shape discrimination and object recognition. In this sense, visual completion is conceptualized not as end in and of itself, but as means to an end.