What's Your Sign?: Eecient Sign Coding for Embedded Wavelet Image Coding

Wavelet transform coeecients are deened by both a magnitude and a sign. While promising algorithms exist for eeciently coding the transform coeecient magnitudes 7, 13, 9, 8, 11], current wavelet image coding algorithms are not eecient at coding the sign of the transform coeecients. It is generally assumed that there is no compression gain to be obtained from entropy coding of the sign. Only recently have some authors begun to investigate this component of wavelet image coding 11, 10, 2]. In this paper, sign coding is examined in detail in the context of an embedded wavelet image coder. It is shown that PSNR improvements up to .7 dB are possible from an eecient modeling and entropy coding of the coeecient signs, combined with a new extrapolation technique, which is used to improve the nal estimate of insigniicant coeecients. These sign coding techniques are applicable to any genre (e.g. zero-tree, context-model) of embedded wavelet image coder.