Spectral and textural classification of single and multiple band digital images

摘要: Abstract Single and multiple band images are classified using supervised algorithms. Two programs, MXTEXT MXMULT, presented that use minimum-distance-to-mean or Bayesian, maximum likelihood algorithms for spectral classification (pattern recognition), further allowing of image texture based on the local variogram surrounding each pixel. Classification can be performed independently information, a combined spectral/textural performed. Combining with information is shown to particular value single radar images. Improved accuracy demonstrated also when variograms cross-variograms used classify information.