Unsupervised hierarchic classification experiment made on raw and compressed data
Abstract
A study on unsupervised classification methods applicable to multispectral and monospectral images is presented. A hierarchic classification of histograms is proposed. The notion of hierarchic regrouping is explained and extended to multispectral images. LANDSAT images and DEADALUS images were processed and the original data and the data compressed in 16 and 32 classes are presented. It is shown that the utilization of compressed data tends to improve the images by sorting out the dominant classes.
- Publication:
-
NASA STI/Recon Technical Report N
- Pub Date:
- January 1982
- Bibcode:
- 1982STIN...8313436L
- Keywords:
-
- Classifications;
- Data Compression;
- Histograms;
- Image Processing;
- Multispectral Photography;
- Pattern Recognition;
- Cluster Analysis;
- Hierarchies;
- Imaging Techniques;
- Infrared Scanners;
- Landsat Satellites;
- Instrumentation and Photography