Multifractal estimation for remote sensing image segmentation

Multifractal analysis can successfully characterize the roughness and self-similarity of textural images. But most popular methods produce less accurate results. In this paper, a novel multifractal estimation algorithm based on mathematical morphology is proposed and a set of new multifractal features, namely the local morphological multifractal exponents (LNME) is defined. A series of cubic Structure Elements (SE) and iterative morphological operations are utilized so that the computational complexity of the new approach can be tremendously reduced. A quadtree-based multilevel segmentation algorithm is also developed to efficiently apply the presented multifractal features to image segmentation. Both die proposed approach and the box-counting based methods have been assessed on real remote sensing images. The comparison results demonstrate that the morphological multifractal estimation can differentiate texture images more effectively and provide a more robust segmentation result.