A fast near-optimum block truncation coding method using a truncated K-means algorithm and inter-block correlation

Block truncation coding (BTC) is an efficient image compression method which finds application in diverse fields. The basic problem can be viewed as a two-level quantization process. However, efficient ways for optimal threshold determination have not been discovered so far. We propose a fast BTC algorithm based on a truncated K-means algorithm, utilizing the image inter-block correlation and the fact that K-means algorithm converges very fast in this case. This produces near-optimum solution with significantly improved speed over other methods. Simulation results confirm such advantages.