Wavelet-based L-infinite scalable coding.pdf

Wavelet-Based L-infinite Scalable Coding A. Alecu, A. Munteanu, P. Schelkens, J. Cornelis and S. Dewitte Authors’ affiliations: A. Alecu, A. Munteanu, P. Schelkens, J. Cornelis and S. Dewitte (Vrije Universiteit Brussel, Dept. of Electronics and Information Processing, Pleinlaan 2, 1050 Brussels, Belgium, aalecu@etro.vub.ac.be) S. Dewitte (Royal Meteorological Institute of Belgium, Ringlaan 3, 1180 Brussels, Belgium) Abstract: A new wavelet-based embedded image coding technique is proposed. For any desired L∞ -distortion bound, the bit stream can be truncated at a corresponding bit-rate, so that the target upper bound on the absolute values of the pixels of the reconstruction error image is guaranteed. The algorithm outperforms embedded 2L -oriented wavelet coders in terms of maximum absolute error. Introduction: Some applications in remote sensing and medical imaging require compression schemes that guarantee constraints on the local reconstruction (pixel) error - the MAXimum Absolute Difference, MAXAD, between the pixel values in the original and the reconstructed image - rather than on the average error. For these cases, we propose a new wavelet-based L∞ - scalable image coding technique that guarantees the required MAXAD, while retaining the coding performances of state-of-the-art wavelet-based 2L -oriented coders [1, 2]. Several methods have been proposed in literature for L∞ - distortion constrained compression [3-5], all targeting a set of fixed reconstruction error thresholds. While it is possible to transmit different images so as to meet the specific MAXAD requirement for each user, such an approach is highly redundant and time-consuming. Our scheme supports local storage of a single embedded bit stream going up to near-lossless reconstruction, which can be truncated at certain bit-rates corresponding to user-defined guaranteed MAXAD values. Distortion-Rate Optimization: The lifting-based wavelet trans