A MODIFIED OCTREE COLOR QUANTIZATION ALGORITHM

Color quantization is widely exploited for many applications. Studying the traditional octree color quantization algorithm (OCQ) found that it is efficient for color image quantization except it needs huge calculating time and computer memory. OCQ adopts a bottom-up pruning process, i.e. beginning with leaf nodes in a complete octree, so to avoid the large numbers of leaves is the key of speed up processing and save memory. A modified octree color quantization algorithm (MOCQ) was proposed to answer the above questions. MOCQ adopted a bidirectional mechanism of first up-bottom comparing then bottom-up pruning directly to improve processing speed, limited the depth of octree to 4 to save memory, and used an error diffusion technology to preserve image quality. The ending conditions for the algorithm could be found out automatically. Experimental results show that this algorithm has rather good performance with high speed and low complexity for real-time and high quantization quality applications.