Thirty-two classes compression experiment: Supervised classification and texture experiment
Abstract
The modification of a clustering algorithm used to compress DAEDULUS images to 16 or 32 classes, followed by a comparison of the results of a supervised classification of the compressed data in 16 or 32 classes, and the evaluation of a Matra algorithm in a compression to 16 or 32 classes of LANDSAT images are discussed. Contingency tables and the detailed parameters used in the experiment are presented. It is concluded that there is no significant difference between the compression to 16 or 32 classes. The Bemzecri distances are smaller in the case of supervised classification. The texture charts of the LANDSAT images show only small differences between the original and the compressed images.
- Publication:
-
NASA STI/Recon Technical Report N
- Pub Date:
- 1982
- Bibcode:
- 1982STIN...8313435L
- Keywords:
-
- Classifications;
- Cluster Analysis;
- Data Compression;
- Image Processing;
- Textures;
- Aerial Reconnaissance;
- Clumps;
- Comparison;
- Iran;
- Landsat Satellites;
- Multispectral Band Scanners;
- Pattern Recognition;
- Instrumentation and Photography