Attribute trees in image analysis - heuristic matching and learning techniques

As a data structure, a tree is an optimal presentation of hierarchical objects. Many irregular and dynamical phenomena studied for example in biology, medical sciences, meteorology, and geomorphology can be modelled as a tree. In addition, objects initially modelled as a graph can sometimes be transformed to a tree, say to a minimum spanning tree. This paper presents new techniques for indexing, matching, and generalizing rooted unordered attribute trees. The proposed matching scheme is based on dividing the tree recursively into subtrees. The subtrees are matched according to topological indices which have been calculated in advance using linear updating rules. The feasibility of the suggested methods is illustrated with experiments on real data extracted from remote sensing imagery