A very low-complexity multi-resolution prediction-based wavelet transform method for medical image compression

Wavelet based lossless compression techniques have been popular for medical image compression due to a number of features, like multi-resolution representation, progressive transmission and high compression ratios. As decoding time is of paramount importance in medical applications, low complexity wavelets would be preferred for fast decoding and retrieval of data from picture archiving and communications systems (PACS) enabling quicker diagnosis and higher productivity of the physician. We propose a novel image compression system that claims extremely low complexity, in fact lower than the Haar wavelet, and at the same time providing higher compression ratios. The high pixel-to-pixel correlation inherent in medical images is first exploited by the application of differential pulse code modulation (DPCM) followed by a modified version of the Haar wavelet applied in an incomplete fashion. We report extensive results (first-order entropy estimates) on a large database of medical images.