Accession Number : AD0285794

Title :   FOURIER-TRANSFORM METHODS FOR PATTERN RECOGNITION

Corporate Author : AIR FORCE CAMBRIDGE RESEARCH LABS L G HANSCOM FIELD MASS

Personal Author(s) : JONES,C.M. ; KELLAHER,J.F.

Report Date : JUL 1962

Pagination or Media Count : 1

Abstract : THE USE OF FOURIER TRANSFORM METHODS FOR CLASSIFYING IMAGES IS INVESTIGATED IN SOME DETAIL. IT IS CONCLUDED THAT THE AMPLITUDES OF FREQUENCY COMPONENTS OF THE WAVEFORM OF A SCANNED IMAGE OFFER A USEFUL DESCRIPTION METHOD. TOPICS INCLUDE A DISCUSSION OF THE FACTORS AFFECTING CLASSIFICATION, METHODS FOR REPRESENTING PATTERNS, COMPARISON OF RELATIVE COMPLEXITIES OF THE TIME AND FREQUENCY REPRESENTATIONS, DEFINITION OF WELL-CONSTRUCTED OF INTERESTING PATTERNS, EFFECTS OF VARYING SCANNING MODES, SEPARABILITY OF PATTERNS REPRESENTED BY THE POWER DENSITY SPECTRUM, AND DESIGN OF A CONCEPTUAL IMAGE CLASSIFICATION SYSTEM

Descriptors :   *FOURIER ANALYSIS, *READING MACHINES, ARTIFICIAL INTELLIGENCE, CLASSIFICATION, COLLECTING METHODS, IDENTIFICATION, PATTERN RECOGNITION, STATISTICAL PROCESSES

Distribution Statement : APPROVED FOR PUBLIC RELEASE