Vision architecture for motion analysis

Summary form only given, as follows. An examination is made of both low-level and higher level vision architectures for motion analysis. To improve the low-level motion information processing, the need to investigate specific issues related to motion detection mechanisms is argued. These issues are: the use of inhibition inputs, the inhibitory mechanism in motion, and the resolution increase. A structure is suggested for higher level processing which integrates different features of the image in addition to motion attributes. When the implementation of a computer vision system is considered, the mapping of these biologically inspired architectures and algorithms into existing parallel computer architectures is important.<>