Efficient lossless coding of medical image volumes using reversible integer wavelet transforms

A novel lossless medical image compression algorithm based on three-dimensional integer wavelet transforms and zerotree coding is presented. The EZW algorithm is extended to three dimensions and context-based adaptive arithmetic coding is used to improve its performance. The algorithm (3-D CB-EZW) efficiently encodes image volumes by exploiting the dependencies in all three dimensions, while enabling lossy and lossless compression from the same bitstream. Results on lossless compression of CT and MR images are presented, and compared to other lossless compression algorithms. The progressive performance of the 3-D CB-EZW algorithm is also compared to other lossy progressive coding algorithms. For representative images, the 3-D CB-EZW algorithm produced an average of 14% and 20% decrease in compressed file sizes for CT and MR images, respectively, compared to the best available 2-D lossless compression techniques.