LOSSLESS COMPRESSION OF MEDICAL IMAGES BY PREDICTION AND CLASSIFICATION

A lossless image compression algorithm based on a prediction and class ification scheme is presented. The algorithm decomposes an image into four subimages by subsampling pixels at even and odd (both row and col umn) locations. Because the four subimages have strong correlations to one another, one of them is used as a reference in predicting the oth er three, and the resulting differences between the predicted subimage s and the original subimages are encoded. Even though these difference s are decorrelated and tend to be random, there is still a relatively large correlation left between the estimated differences and the subim age used in prediction. For example, the estimated differences tend to be large in a high-detailed region, where pixel values change rapidly , whereas the differences are small in a low-detailed region, where pi xel values change smoothly and slowly. This redundancy is exploited by dividing the estimated differences into subsets based on a slope chan ge in the subimage used in prediction. The performance of the proposed algorithm using two different predictors, linear interpolation and th ird-order polynomial interpolation, is compared with that of the hiera rchical interpolation (HINT) scheme and Fourier transform interpolatio n by measuring the first-order entropies of the estimated differences. With a third-order polynomial interpolation and division into two sub sets, an average entropy of 3.1 bits/pixel is achieved for the three p redicted difference subimages of the 12 bits/pixel x-ray computed tomo graphy images. It is about 0.86 bits/pixel lower in the first-order en tropy than the HINT for the three subimages.