Binary partition tree as an efficient representation for image processing, segmentation, and information retrieval

摘要

This paper discusses the interest of binary partitionudtrees as a region-oriented image representation. Binary partitionudtrees concentrate in a compact and structured representation a setudof meaningful regions that can be extracted from an image. Theyudoffer a multiscale representation of the image and define a translationudinvariant 2-connectivity rule among regions. As shown in thisudpaper, this representation can be used for a large number of processingudgoals such as filtering, segmentation, information retrievaludand visual browsing. Furthermore, the processing of the tree representationudleads to very efficient algorithms. Finally, for some applications,udit may be interesting to compute the binary partition treeudonce and to store it for subsequent use for various applications. Inudthis context, the last section of the paper will show that the amountudof bits necessary to encode a binary partition tree remains moderate.