In addition to spectral features, texture is an important spatial feature used in identifying objects or regions of interest in an image. Although texture is relatively easy for human observers to recognize and describe in empirical terms, it has been extremely refractory to precise definition and analysis by digital computers. This paper describes a procedure for extracting some easily computable features for the texture of blocks of digital image data and illustrates the applications of combined textural (spatial) and spectral features for identifying the land use categories of blocks of ERTS MSS (Earth Resources Technology Satellite Multi Spectral Scanner) data. The land use classification algorithm based on textural and spectral features was developed and tested using 614 image blocks of 64*64 resolution cells derived from an ERTS image over the Monterrey Bay area of the California coast line. The algorithm was applied on a training set of 314 blocks and tested on a set of 310 blocks. The overall accuracy of the classifier was found to be 83.5% on seven land use categories
Combined spectral and spatial processing of ERTS imagery data
Kombinierte spektrale und raeumliche Verarbeitung von ERTS-Abbildungsdaten
Remote Sensing Environ. (USA) ; 3 , 1 ; 3-13
1974
, 20 Quellen
Article (Journal)
English
Spectral and textural processing of ERTS imagery
NTRS | 1974
|Crop identification using ERTS imagery
NTRS | 1973
|Pseudocolor transformation of ERTS imagery
NTRS | 1973
|