A Comparative Study of Segmentation Algorithms for FLIR Images
Abstract
A comparative study of FLIR segmentation algorithms has been conducted in cooperation with Westinghouse Defense Systems Division. In the Maryland portion of the study, four techniques (two and three-class relaxation, pyramid linking, and superspike) were tested on a Westinghouse-supplied database of 51 images obtained from NVL and other sources. (Two other techniques, superslice and pyramid sport detection, were rejected after preliminary studies. ) The best technique, superspike, extracted regions corresponding to over 88% of the targets, and had a false alarm rate of 1.6 false regions per true target.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1981
- Accession Number
- ADA124809
Entities
People
- Azriel Rosenfeld
- Cheng-ye Wang
- Leslie J. Kitchen
- Ralph L. Hartley
Organizations
- University of Maryland