Algorithms for speedy visual recognition and classification of patterns formed on rectangular imaging sensors

We present a survey of conceptual and technological challenges facing the computer vision practitioners. The real-time tasks of image interpretation, pattern clustering, recognition and classification necessitate quick analysis of registered patterns. Current algorithms and technological approaches are slow and rigid. Within the evolutionary context, having performed the simulation of human visual cortex, we present a novel, biologically inspired solution to these problems, with aim to accelerate processing. Our massively parallel algorithms are applicable to patterns recorded on rectangular imaging arrays of arbitrary sizes and resolutions. The systems we propose are capable of continuous learning, as well as of selective forgetting less important facts, and so to maintain their adaptation to an environment evolving sufficiently slowly. A case study of visual signature recognition illustrates the presented ideas and concepts.