Accession Number:

ADA043495

Title:

A New Approach to Structure Preserving Feature Extraction.

Descriptive Note:

Interim rept.,

Corporate Author:

RICE UNIV HOUSTON TEX DEPT OF ELECTRICAL ENGINEERING

Personal Author(s):

• Starks,Scott A.
• De Figueiredo,Rui J. P.

Report Date:

1977-02-15

Pagination or Media Count:

9.0

Abstract:

This paper presents an approach to nonlinear feature extraction based on certain graph theoretical considerations such as the minimal spanning tree, maximally complete subgraphs, inconsistent edges and diameter edges and topological considerations such as interpoint distance measures. After appropriate introductory sections, the feature extraction algorithm is developed. The algorithm is hierarchical in nature and offers considerable savings in terms of computer computation and storage requirements. An outline of the computer procedure is also included. Author

Descriptors:

• *COMPUTER PROGRAMMING
• *PATTERN RECOGNITION
• ALGORITHMS
• DATA PROCESSING
• GRAPHS
• SEQUENTIAL ANALYSIS
• NONLINEAR SYSTEMS
• CLUSTERING
• TOPOLOGY
• HEURISTIC METHODS
• DATA STORAGE SYSTEMS
• FLOW CHARTING
• DATA COMPRESSION
• HIERARCHIES
• MAPPING(TRANSFORMATIONS)

Subject Categories:

• Statistics and Probability
• Computer Programming and Software
• Computer Hardware
• Bionics

Distribution Statement:

APPROVED FOR PUBLIC RELEASE