|
|
|
“A Digital Optical Cellular Image Processor (DOCIP): Theory, Architecture and Implementation”
by Kung-Shiuh Huang
November 1988
Is there a simple unified complete theory of parallel binary digital image processing? Can this theory suggest new parallel algorithms and architectures for image processing and numerical computation? Can these algorithms and architectures be implemented by optical computing techniques? This thesis attempts to answer these questions.
A binary image algebra (BIA), built from five elementary images and three fundamental operations, serves as a unified theory of parallel binary image processing and a spatial logic of parallel optical computing. It also leads to a formal parallel language approach to the design of parallel binary image processing and parallel binary arithmetic algorithms. Digital optical cellular image processors (DOCIPs), based on cellular automata and cellular logic architectures, implement parallel algorithms of BIA efficiently. An algebraic structure provides a link between the algorithms of BIA and architectures of DOCIP.
Optical computing suggests an efficient and high-speed implementation of the DOCIP architectures because of its inherent parallelism and 3-D global interconnection capabilities. The use of optical interconnections permits a two-dimensional cellular hypercube (DOCIP-hypercube) topology to be implemented without paying a large penalty in chip area (the cellular hypercube interconnections are space-invariant which implies a low hologram complexity); it also enables images to be input to and output from the machine in parallel. A computer-controlled system has been constructed to fabricate multi-facet interconnection holograms for 3-D optical circuits. A prototype DOCIP system has been implemented to demonstrate the concept of the DOCIP architecture. Experimental results are presented.