Improved Noise Suppression for Multiband Imagery.
Abstract
This research has developed and tested an approach to improvement of noise degraded multiband imagery based on a class dependent Karhunen-Loeve (K-L) transformation and transform domain spatial filtering. The technique is attractive because of its tendency to preserve the high spatial frequency information (e.g., edges) in the picture as compared with conventional minimum mean square error linear filtering of the original bands. Multiband images (especially color and multispectral characteristically exhibit significant class structure, and within a particular class show high spectral correlations. The filtering technique developed during this research exploits both the class structure and the intraclass correlation to achieve noise suppression. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1980
- Accession Number
- ADA091815
Entities
People
- B. J. Gaitley
- D A Anderson
- P. J. Ready