Accession Number : ADA297910
Title : Hardware Implementation of a Desktop Supercomputer for High Performance Image Processing. Feasibility Study of a Time Multiplexing Approach Using a 4x4 CNN Laboratory Prototype.
Descriptive Note : Technical rept. 1 May-1 Aug 95,
Corporate Author : TEXAS A AND M UNIV COLLEGE STATION DEPT OF ELECTRICAL ENGINEERING
Personal Author(s) : De Gyvez, Jose P.
PDF Url : ADA297910
Report Date : 14 AUG 1995
Pagination or Media Count : 16
Abstract : The state of the art work has concentrated on VLSI CNN implementations without really addressing the "system level". While efficient implementations have been reported, no reports have been presented on the use of these implementations for processing large complex images. This report presents a feasibility study to use our time multiplexing scheme based on a CNN prototype. The laboratory prototype consists of a 4x4 CNN analog discrete implementation serially linked to a PC. A simple graphical interface has been built to be able to read bitmap images of any size and to send them to the CNN. The results are encouraging as they demonstrate the potential of the multiplexing approach applied to complex black and white images. These images consist of two aerial views of 256x256 pixels processed only with the 4x4 CNN.
Descriptors : *IMAGE PROCESSING, *SUPERCOMPUTERS, NEURAL NETS, STATE OF THE ART, TOOLS, VERY LARGE SCALE INTEGRATION, PROTOTYPES, FEASIBILITY STUDIES, LABORATORY PROCEDURES.
Subject Categories : Electrical and Electronic Equipment
Distribution Statement : APPROVED FOR PUBLIC RELEASE