Mode based K-means algorithm with residual vector quantization for compressing images

Abstract

Image compression plays a vital role in many online applications like Video Conferencing, High Definition Television, Satellite Communication and other applications that demand fast and massive transmission of images. In this paper, we propose a Mode based K-means method that combines K-Means and Residual Vector Quantization(RVQ) for compressing images. Three processes are involved in this approach; Partitioning and Clustering, Pruning and Construction of Master codebook and Residual vector Quantization. Extensive experiments show that this method obtains a fast solution with better compression rate and comparable PSNR than conventional K-Means algorithm. © 2011 Springer-Verlag Berlin Heidelberg.