CLASSIFICATION AND SEGMENTATION BY UNSUPERVISED K-MEANS TECHNIQUE BASED ON DECOMPOSITION OF SINGLE-BAND REMOTELY SENSED IMAGES USING WAVELET TRANSFORMATION
S. M. Ali
Remote Sensing Unit- College of Science- University of Baghdad IRAQ- Baghdad- Al-Jaderyia
An adaptive unsupervised K-means classification and segmentation techniques are introduced, based on decomposition of single-band remotely sensed images, using stationary wavelet- transform. Images by our classification technique are decomposed into 4 sub-bands using two-level stationary wavelet- transform. For each image point, a 4-dimensional vector is constructed whose components represent the same location values at the 4-wavelet sub-bands. K-means unsupervised classification method is used to perform the classification process. Image segmentation, based on color or gray slicing method is also presented..
Keywords, Classification Method, Segmentation Method, Unsupervised Classification and Segmentation, K-Means Classification, Wavelet Classification Method, Band decomposition classification, Remote sense data classification and Segmentation.