Video compression using quadtree segmentation and component quantization

The use of component quantizers to improve the coding gain of quadtree segmentation-based video coders is investigated. Quadtree segmentation is a hierarchical spatial data decomposition technique that represents different spatial resolutions at different levels of the tree. For video coding applications, the sample averages associated with each node region of the quadtree are quantized and transmitted. In the present work, the sample average distribution corresponding to each tree level is analyzed. The statistical analysis shows that the sample distribution differs from level to level. Thus, it is proposed that a component quantizer be used for the sample quantization. Its superiority over a single-source quantizer is demonstrated.<>