SBIR-STTR Award

The proposed innovation is real-time hardware for image processing, image understanding, and machine vision that will achieve significant performance increases over current systems by implementing the reduced set of operations required to support aspecific set of algorithms. These algorithms are the "generic neighborhood operations" of Koenderink and van Doorn whichprovide a framework for describing the local structure of objects in images at a point.The goal of Phase I of this project is to perform the technical research and detailed engineering design of a system to implementthis theory of generic neighborhood operators in real-time. This project is concerned with performing the engineering design for real-time hardware for image processing and image understanding.In its final implementation, it will achieve orders of magnitudeperformance increases over currently available systems byimplementing only a reduced set of operations required to supporta specific set of algorithms. In particular, in its final configuration, it may process at up to 138 G operation per second.Commercial Applications:(1) Robotics: vision systems inunconstrained environments; (2) Medical images: mammography, papsmears, chest x-rays; (3) Security: face recognition, fingerprintidentification; (4) Automated inspection: unconstrained objectrecognition; (5) Office automation: OCR, forms recognition; and(6) General computer vision research.

The Proposed Innovation is real-time hardware for image processing, image understanding and machine vision that will achieve significant performance increases over current systems by implementing a set of operations required to support z specificset of algorithms. These algorithms are the generic neighborhoodoperations which produce multiple, parallel partial derivativeimages or Whet bundles". These provide a framework fat describingthe local structure of objects in images at a point.The goal of Phase II of this project is to design arc construct 2prototype Generic Neighborhood Operator and Scale Space Processor(GNOSSP) which will process three color input images and produce225 partial derivative images - the jet bundle. This GNOSSP willfunction as a first stage for a subsequent image vector processorwhich will analyze these jet bundles. It will perform thesealgorithms at real-time video frame rates with a 640 x 480 imageformat. It will make use of commercially available programmablegate array integrated circuits. In its final implementation, itwill achieve orders of magnitude performance increases overcurrently available systems by implementing only the operationsrequired to support the specific set of algorithms.Potential commercial applications of the research:1 Robotics:vision systems in unconstrained environments 2 Medical images:mammography, pap smears, chest x-rays 3 Security: facerecognition, fingerprint identification 4 Automated inspection:unconstrained object recognition 5 Office automation: OCR, formsrecognition 6 General computer vision research