Analyzing information content of MR images

Abstract

The information content of magnetic resonance images is analyzed for the purpose of achieving more compression using the existing algorithms. It is found that the first five bits of all images contain no information and therefore can be totally compressed. The experiments also showed that, by segmenting and transforming the image, the entropy and therefore the compression rate can change. A vertical (along z axis) segmentation method is presented which can improve the compression rate by 10%. Other types of segmentation method (along x or y axes) can also be studied. The transformation method used was a simple differential coding method which resulted in almost no improvement in compression. This was an interesting observation, since this same transformation can drastically improve the compression of other types of data such as satellite imagery.<>