Accession Number : AD0863596

Title :   Multi-Image Pattern Recognition: Ideas and Results.

Descriptive Note : Technical rept.,

Corporate Author : KANSAS UNIV LAWRENCE CENTER FOR RESEARCH IN ENGINEERING SCIENCE

Personal Author(s) : Haralick, Robert M.

Report Date : 1969

Pagination or Media Count : 220

Abstract : A tutorial introduction to some of the pattern recognition problems is presented in a unified perspective. Special concentrated effort is given to pattern recognition techniques for use with sequential data especially multi-image data sets. Some of these techniques are experimentally tested on a photographic data set. The following are among the original pattern recognition techniques presented; identifying a data set with a measure of association, spatial quantizing, measurement space clustering, spatial clustering, texture analysis, registering and congruencing images. The spatial quantizing and spatial clustering techniques were tested with some imagery from Michigan's 12 channel scanner system. The spatial quantizing procedure produced a better quality quantized image than the linearly quantized image. The spatial clustering procedure produced a map of the natural homogeneous classes of environmental objects seen like the sensor sees them. (Author)

Descriptors :   (*PATTERN RECOGNITION, REVIEWS), PHOTOGRAPHIC IMAGES, DECISION THEORY, DETECTORS, CORRELATION TECHNIQUES, CLASSIFICATION, THESES.

Subject Categories : Psychology
      Cybernetics
      Bionics

Distribution Statement : APPROVED FOR PUBLIC RELEASE