Image data compression using multiple bases representation

Abstract

Digitized images contain huge amounts of information which strain, or exceed, the capacity for their real-time processing, storage, and retrieval. Various compression techniques have been developed to reduce the amount of data necessary for representation. We report on a hybrid image data compression procedure based on a multiple bases representation. The multiple bases representation technique described herein utilizes advantages of transform coding, vector quantization, and predictive coding, while aiming to circumvent the associated disadvantages of each. Preliminary results indicate that this procedure can outperform conventional compression methods, and yield high compression ratios while avoiding prohibitive computational complexity.