Computers and Biomedical Research 32, 449–469 (1999) Article ID cbmr.1999.1523, available online at http://www.idealibrary.com on Best Parameters Selection for Wavelet Packet-Based Compression of Magnetic Resonance Images A. N. Abu-Rezq, A. S. Tolba, 1 G. A. Khuwaja, and S. G. Foda* Physics Department, Kuwait University, P.O. Box 5969, 13060 Safat, Kuwait and; *Electrical and Computer Engineering Department; Kuwait University, Kuwait Received January 8, 1999 Transmission of compressed medical images is becoming a vital tool in telemedicine. Thus new methods are needed for efficient image compression. This study discovers the best design parameters for a data compression scheme applied to digital magnetic resonance (MR) images. The proposed technique aims at reducing the transmission cost while preserving the diagnostic information. By selecting the wavelet packet’s filters, decomposition level, and subbands that are better adapted to the frequency characteristics of the image, one may achieve better image representation in the sense of lower entropy or minimal distortion. Experimental results show that the selection of the best parameters has a dramatic effect on the data compression rate of MR images. In all cases, decomposition at three or four levels with the Coiflet 5 wavelet (Coif 5) results in better compression performance than the other wavelets. Image resolution is found to have a remarkable effect on the compression rate.  1999 Academic Press Key Words: wavelet packets; best parameter selection; magnetic resonance imaging; compres- sion; telemedicine. 1. INTRODUCTION The unacceptability of artifacts when making medical diagnosis has limited the wide use of image compression by the medical community. However, the imperative for compression grows as medical centers use gigabytes or terabytes of disk space per year. Transmission of medical images is becoming a vital tool in telemedicine. Lossless compression is artifact free. Also high-quality lossy compression, with its higher compression rate, could become useful for archiving images. Wavelet packets provide true lossy compression without many of the artifacts of block-based transform compressors. Reconstructed digital images must preserve the necessary information content of the original images to avoid diagnostic errors. The perfor- mance of MR image compression algorithm is measured by its ability to minimize entropy while preserving all clinically significant features. At high and medium 1 To whom correspondence should be addressed. E-mail: tolba@kuc01.kuniv.edu.kw. 449 0010-4809/99 $30.00 Copyright  1999 by Academic Press All rights of reproduction in any form reserved.