Cellular Pixel Array for Real Time Image Processing

In spite of their high speed and precision, contemporary computing platforms fail to deliver the computational density required for real- time image processing towards the realization of truly intelligent and autonomous behavior in today's weapons systems and computers. And yet the simplest of living systems are able to perceive and interpret their environment effortlessly, using only a conglomerate of slow and inaccurate neurons. Moving from this observation, this project proposes to develop a VLSI image processing chip that implements the local connectivity patterns of the visual cortex in silicon to perform local spatio-temporal image processing. The resulting chips will be a general-purpose tool that can be programmed to perform many types of iconic manipulations, including convolution, logic, and morphological operations in the spatial and temporal domains. The Phase I effort will be geared towards proof of concept through system simulation and produce the detailed conceptual designs of the prototype chips to be manufactured and demonstrated during Phase II.