Fast color quantization using weighted sort-means clustering

摘要: Color quantization is an important operation with numerous applications in graphics and image processing. Most methods are essentially based on data clustering algorithms. However, despite its popularity as a general purpose algorithm, K-means has not received much respect the color literature because of high computational requirements sensitivity to initialization. In this paper, fast method presented. The involves several modifications conventional (batch) including reduction, sample weighting, use triangle inequality speed up nearest-neighbor search. Experiments diverse set images demonstrate that, proposed modifications, becomes very competitive state-of-the-art terms both effectiveness efficiency.