Multi-resolution segmentation and shape analysis for remote sensing image classification

We present an approach for classification of remotely sensed imagery using spatial information extracted from multi-resolution approximations. The wavelet transform is used to obtain multiple representations of an image at different resolutions to capture different details inherently found in different structures. Then, pixels at each resolution are grouped into contiguous regions using clustering and mathematical morphology-based segmentation algorithms. The resulting regions are modeled using the statistical summaries of their spectral, textural and shape properties. These models are used to cluster the regions, and the cluster memberships assigned to each region in multiple resolution levels are used to classify the corresponding pixels into land cover/land use categories. Final classification is done using decision tree classifiers. Experiments with two ground truth data sets show the effectiveness of the proposed approach over traditional techniques that do not make strong use of region-based spatial information.