High-order entropy coding of medical image data using different binary-decomposed representations

Information theory indicates that coding efficiency can be improved by utilizing high-order coding (HOEC). However, serious implementation difficulties limit the practical value of HOEC for grayscale image compression. In this paper we present a new approach, called binary-decomposed high-order entropy coding, that signifucantly reduces the complexity of the implementation and increases the accuracy in estimating the statistical model. In this appraoch a grayscale image is first decomposed into a group of binary sub-images. When HOEC is applied to these sub-images instead of the original image, the subsequent coding is made simpler and more accurate statistically. We apply this coding technique in lossless compression of medical images and imaging data, and demonstrate that the performance advantage of this approach is significant.