Linear prediction image coding using iterated function systems

Abstract

This paper presents a hybrid system to speed up image fractal encoding. The coding scheme, LP–IFS, consists of linear prediction (LP) and Iterated Functions Systems (IFS) applied in cascade on the image. The LP process employs a 2D auto-regressive model to estimate parameters for each block in the image partition; IFS are then used instead of adaptive quantizers to encode linear prediction errors. The stability of the resulting coding scheme is assured, since both LP and IFS are stable systems. The experiments performed have shown that LP–IFS can achieve very low bit-rates (BR) with good subjective and objective quality. Moreover, comparative studies based on extensive computer simulations have demonstrated that LP–IFS can rival standard IFS-based techniques in terms of BR and peak signal-to-noise ratio for high compression ratio and with respect to computing time.