Segmentation of multimodal images based on hierarchies of partitions

Abstract

Hierarchies of partitions are widely used in the context of image segmentation, but when it comes to multimodal images, the fusion of multiple hierarchies remains a challenge. Recently, braids of partitions have been proposed as a possible solution to this issue, but have never been implemented in a practical case. In this paper, we propose a new methodology to achieve multimodal segmentation based on this notion of braids of partitions. We apply this new method in a practical example, namely the segmentation of hyperspectral and LiDAR data. Obtained results confirm the potential of the proposed method.