Fuzzy clustering of textured SAR images based on a fractal dimension feature

Texture is an important feature to improve image segmentation. This paper describes a wavelet-based fuzzy clustering procedure which receives as input both the original image and a texture map based on a fractal model, derived from the wavelet transform of the image itself. This preliminary evaluation has shown the potential usefulness of the fractal signature for improving texture discrimination and enhancing segmentation results, and suggests this multi-scale approach for remote sensing image texture analysis.