Attentive visual recognition

Vision is an active process where behaviorally important information is selectively gathered. We present a model of visual recognition in which the high-resolution fovea is deployed to interesting regions by selective attention processes. The raw image is initially represented by a space-variant complex-cell map which is processed by a decision system that attempts to recognize the image given partial information. If the available information is insufficient to support recognition, an attentive foveation system is engaged and determines the subsequent image region of interest. The next foveation is based on both bottom-up information from the image and top-down information from the set of stored models, and corresponds to the process of selective attention within the model. The foveation process continues until the object is recognized. The model is applied to the task of face recognition where simulations illustrate its behavior