Compressed quadtree with content-addressable memory

The paper is concerned with the development of linear quadtree structure for image coding with the objective of making it more compact and suitable for content addressable memory (CAM) implementation. Linear quadtree represents an image by a collection of leaf nodes which are represented by locational nodes. The author aims at reducing the number of locational codes by making use of the coherence of regions. The underlying idea is to extend the use of “dont care” state in the CAM-based linear quadtree to exploit more spatial coherence. The author proposes a new hierarchical data structure, named as quad-quadtree, which is based on the recursive subdivision of an image into 16 equal-sized sub-images until each is homogeneous or until a sufficiently fine resolution is reached. The quad-quadtree differs from other quadtree structures in that it reduces the number of codes by using rearranged labelling order and merging homogeneous regions. Experimental results obtained on real data demonstrate its space efficiency and usefulness