Effective simple criteria for feature subset selection
Abstract
A feature type or groups of feature types will often define more features than can be used in the classification of regions in an image. Thus it is desirable to select a subset of the available features which are optimum in the sense of a defined criteria. A number of previously defined selection criteria utilize the data found in the within-class scatter matrix and the between-class scatter matrix. This paper describes two simpler selection criteria which also use data contained in the within-class and between-class scatter matrices. The criteria are defined and the rationale behind their selection indicated. The effectiveness of the simple feature selection criteria in selecting features is evaluated with experimental classification results.
- Publication:
-
IGARSS 1987 - International Geoscience and Remote Sensing Symposium
- Pub Date:
- 1987
- Bibcode:
- 1987grs..conf..679C
- Keywords:
-
- Image Analysis;
- Pattern Recognition;
- Photointerpretation;
- Target Acquisition;
- Classifying;
- Fourier Transformation;
- Image Processing;
- Optimization;
- S Matrix Theory;
- Instrumentation and Photography