Algorithms and Hardware Technology for Image Recognition
Abstract
Image models have been developed to propose thresholds, classify pixels and predict operator response to both signal and noise. A comprehensive approach to target cueing in FLIR imagery has been developed, based on simple but powerful heuristics. 'Superslice', the algorithm embodying this approach, has been tested on two data bases and has successfully extracted object regions while accepting few noise regions. The target regions are discriminable from noise; however, they are difficult to identify reliably as to type.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 30, 1977
- Accession Number
- ADA041906
Entities
People
- Azriel Rosenfeld
- David L. Milgram
Organizations
- University of Maryland