Fuzzy Similarity Measure and its Application to High Resolution Colour Remote Sensing Image Processing

The focus for the study in this thesis is placed on developing basic algorithms and tools for high-resolution colour remote sensing image processing tasks such as colour morphology, multivariate clustering, and multivariate filtering. First, the fuzzy similarity measure (FSM) among vectors in a vector space is introduced. This measure is based on two assumptions for the relationship among vectors: short-range ordering and fuzzification. Second, based on the FSM, the colour morphology, multivariate fuzzy clustering, and multivariate filtering are defined. The performances of all proposed methods will be evaluated numerically and subjectively. Third, this study also places more emphases on solving some applied problems related to recognizing colour edges, detecting and extracting complex road network and building rooftops, and reducing noise in high-resolution remote sensing images such as QuickBird, Ikonos, and aerial images. The results obtained in the study demonstrate the effectiveness and efficacy of the FMS and the proposed methods.