Smart Analog Vision Chip with Spatio Temporal filtering

During the last decade a new generation of biologically motivated imaging sensors have been under development at several universities. This research initiative is spear headed by ONR funded research at Caltech. These devices include co-located sensing elements and pixel level data processing elements. The data processing circuits perform early levels of image processing tasks. Whereas in existing machine vision systems (based on a CCD camera, A/D converter and a programmable computer) 60-80% of computational power is devoted to early levels of image processing, the new generation of sensors achieve this is on a single chip. This saving in computational burden can be used to reduce the size and cost of machine vision system. Although, in the development of these new sensors, researchers have focused on extraction of spatial features, early levels of image processing in biological systems also includes the detection of motion fields. It is well known that we rely on this type of computation for survival. The research effort to be undertaken in this SBIR will focus on computation of motion (spatio-temporal information) in a CMOS image sensor. The information generated by such a sensor will be further processed to develop cameras for military and industrial security applications.