Contextual Medical Image Compression using Normalized Wavelet-Transform Coefficients and Prediction

Context-based compression plays a vital role in digital communication systems, since a particular region alone can be preserved using high bit rate and the other regions can be compressed using low bit rate compressions. Such methods are of great interest in tele-radiology applications requiring large storage. This paper presents an enhanced method for compression of medical images using wavelet transformation, normalization, and prediction. The compression method can be tuned to reproduce a good quality image close to the original image for the selected contextual area. Initially, the image undergoes 2D wavelet transform to obtain the approximate and the detailed coefficients. To ease the process of prediction, normalization is done for each sub-band separately, followed by mask-based prediction of the normalized coefficients. Finally, the prediction error coefficients are entropy-encoded using arithmetic coding technique. The proposed algorithm utilizes prediction as well as transformation to achieve a better compression along with good quality. The performance of the proposed system is compared with JPEG2000 and other conventional and contextual compression algorithms. The results show better performance quantitatively and visually.