Improving SPIHT-based Compression of Volumetric Medical Data

Volumetric medical data (CT,MR) are useful tools for diagnostic investigation however their usage may be made diffcult because of the amount of data to store or because of the duration of communication over a limited capacity channel. In order to code such information sources we present a progressive three dimensional image compression algorithm based on zerotree wavelet coder with arithmetic coding. We make use of a 3D separable biorthogonal wavelet transform and we extend the zerotree SPIHT algorithm to three dimensions. Moreover we propose some improvements to the SPIHT encoder in order to obtain a better rate distortion performance without increasing the computational complexity. Finally we propose an efficient context-based adaptive arithmetic coding which eliminates high order redundancy. The results obtained on progressive coding of a test CT volume are better than those presented in recent similar works both for the mean PSNR on the whole volume and for the PSNR homogeneity between various slices.