Using an Octree-Based RAG in Hyper-Irregular Pyramid Segmentation of Texture Volume

Octree Splitting We extend the irregular pyramid segmentation scheme to three dimensional (3D) digital space and we call it hyper-irregular pyramid segmentation. Based on this segmentation scheme, a 3D texture image is split into partitions recursively based on octree structure. The testure features are calculated based on 3D gray level spatial dependency measurement. The octree is subsequently converted into a 3D region adjacency graph (RAG). Each vertex of the graph consists of a texture feature vector of the corresponding partition and each edge represents the neighborhood relationship between two partitions. The hyper-irregular pyramid process is then applied to the graph and finally a segmentation of the three dimensional image is obtained.