A high speed image understanding system

Vision systems, including the human eye, typically start with feature extraction to analyze an image. This is the process of taking a raw image and measuring basic characteristics, for example edges of various orientations, strokes of various thicknesses, end-lines, etc., in order to describe its content with an ‘alphabet’ of basic shapes. This provides a compact representation of the image content that is well suited for interpretation. Objects can be identified from the presence of a few shapes and therefore, from such a representation, a layout analysis of a scene can be done efficiently.