I.8 - Two-Dimensional Maps and Biological Vision: Representing Three-Dimensional Space

Abstract

This chapter discusses two-dimensional spatial transformations of biological and machine vision systems for performing visual tasks such as recognition, path planning, navigation, obstacle avoidance, or tracking. Re-mapping is a very powerful tool for manipulating information contained in a two-dimensional signal such as an image. It has been underexplored by machine vision systems using neural networks. This is unfortunate as these processing methods are easily implemented in neural networks. Biological vision systems are full of examples of two-dimensional re-mapping, at all levels of the neural hardware. This chapter discusses both the strength and necessity of re-mapping for machine vision systems and systems modeling biological vision systems. It reviews the early part of the human visual system concentrating on the possibilities created by a foveated sampling system and multiple functional streams of processing on visual perception. The processes occur early in the visual system and will greatly alter the flow of information to higher centers of the visual cortex. The machine vision system uses many of the transformations found in the human visual system. In vision, it may be critically important that the spatial properties of the image is used as a tool in systems using neural networks.