Hyperspectral Image Segmentation Using a New Spectral Unmixing-Based Binary Partition Tree Representation

Miguel Angel Veganzones (1, 2) , Guillaume Tochon (2) , Mauro Dalla Mura (2) , Antonio J. Plaza (3) , Jocelyn Chanussot (2)

1 GIPSA-CICS - GIPSA - Communication Information and Complex Systems
2 GIPSA-SIGMAPHY - GIPSA - Signal Images Physique
3 Hyperspectral Computing Laboratory (Caceres, Spain)

The Binary Partition Tree (BPT) is a hierarchical region-based representation of an image in a tree structure. BPT allows users to explore the image at different segmentation scales. Often, the tree is pruned to get a more compact representation and so the remaining nodes conform an optimal partition for some given task. Here, we propose a novel BPT construction approach and pruning strategy for hyperspectral images based on spectral unmixing concepts. Linear Spectral Unmixing (LSU) consists of finding the spectral signatures of the materials present in the image (endmembers) and their fractional abundances within each pixel. The proposed methodology exploits the local unmixing of the regions to find the partition achieving a global minimum reconstruction error. Results are presented on real hyperspectral data sets with different contexts and resolutions.

Mots clés: Binary partition trees; segmentation; spectral unmixing; hyperspectral images

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Miguel Angel Veganzones, Guillaume Tochon, Mauro Dalla Mura, Antonio J. Plaza, Jocelyn Chanussot. Hyperspectral Image Segmentation Using a New Spectral Unmixing-Based Binary Partition Tree Representation. IEEE Transactions on Image Processing, 2014, 23 (8), pp.3574 - 3589. ⟨10.1109/TIP.2014.2329767⟩. ⟨hal-01010430⟩