FUZZY CLUSTERINGS OF AERIAL IMAGES AND THEIR EVALUATION BASED ON IMAGE DEPENDENT AREA PARTITION.

Texture analysis of color aerial images has been done based on various Fuzzy clustering algorithms. Each image is given by RGB space (512 multiplied by 512 pixels, 8 bits gray levels, and is fed to an image processing system TOSPIX & DS600/80. Each image is divided into image dependent regions and texture features are calculated for each region of RGB space such as gray-level-average, standard deviation and B. V. Q. (bounded variation quantity) for theta equals 0 degree , 45 degree ,90 degree , 135 degree . Each region is expressed as an element/vector of 18-dimensional vector space. Twelve ( equals 4 multiplied by 3) Fuzzy clustering methods that are F. ISODATA, TN TD, TW with power indices p equals 1. 2, 1. 6, 2. 0 are applied to these vector sets. As a result 12 Fuzzy texture region images are obtained for each aerial image.