Progressive near-lossless coding of medical images

We propose an algorithm intended for compression of medical images, which allows for embedded coding in L/sub /spl infin// sense, i.e., progressive near-lossless as well as lossless image compression. The method is based on a lossy plus near-lossless layered compression scheme and embedded quantization of the difference signal. We show, that this technique allows for a better image quality and compression performance for large tolerance values than algorithms based on predictive coding. The lossy plus near-lossless scheme also allows for image reconstruction with a minimum mean square error (MSE) criterion, while providing a strict control of the maximum absolute difference error. This property is impossible in predictive coding algorithms.