Texture-Tone Feature Extraction and Analysis. Phase II.

Abstract

Since more than 50% of the 17 texture feature variables developed in the Phase I effort are not normally distributed, a non-linear classifed based on stable distribution model was developed to accommodate the skewness characteristics of the texture data and thus improve th hit-rate of the decision maps of the test sets. With this classifer the needed texture feature variables were reduced to 3 from the original 17, resulting in a drastic improvmenet in the processing rate, i.e., a factor of 6 (times faster) has been achieved (from 90 minutes CPU to 15 minutes CPU with IBM 370-158 system). Currently the processing rate for (256 x 256) pixels with the ITEL AS-6 system is less than 4 minutes with FORTRAN programming language. The new classifier was able to improve the hit-rate by about 5 percentage points with an average hit-rate of 90% or higher. In addition, the amount of rejects was also reduced. With a manual method of the selection of training sets, the amount of rejects remains high even though a better classifier was used in the anlaysis. To solve this problem an automatic selection of training sets technique was developed with the intention that the rejects be included in the clusterings derived from randomly selected pixels from the test sets. With this method, the amount of rejects has had been reduced to 3% or less from 20-30 %.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 1979
Accession Number
ADA068817

Entities

People

  • Eugene Klimko
  • Shin-yi Hsu

Organizations

  • Binghamton University

Tags

Communities of Interest

  • Energy and Power Technologies
  • Human Systems
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Air Force
  • Computational Science
  • Data Mining
  • Data Processing
  • Data Science
  • Feature Extraction
  • Image Processing
  • Images
  • Information Processing
  • Information Science
  • Language
  • Machine Learning
  • New York
  • Photographs
  • Photography
  • Statistical Algorithms
  • Test Sets

Fields of Study

  • Computer science

Readers

  • Computer Science.
  • Computer Vision.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference