An improved piecewise approximation algorithm for image compression

摘要

One-dimensional lossy compression schemes, such as piecewise approximation with triggers (PAT) (Walach and Karnin, IEEE Int. Conf. on Acoustics, Speech and Signal Processing, Vol. 1, pp. 529–532, Tokyo, Japan, April 1986) have the advantage of being computationally simple. Several previous papers have examined this algorithm and a variety of changes have been proposed. This paper adds new features and also incorporates some ideas from previous papers to arrive at a modified version: MPAT. The modifications vary parameters for the trigger function and the thresholds. Furthermore, they also introduce new features such as context modelling, different interpolations, and early triggers. These changes add flexibility, decrease the compressed image size, and improve the reconstruction quality, while maintaining a complexity advantage over other algorithms. A complexity count is also performed to quantitatively demonstrate the benefits of this algorithm over alternatives.