Contrast-based adaptive threshold coding for medical images

This paper proposes a contrast-based adaptive threshold coding (CBATC) strategy for medical images under a ´lossy - visually lossless - lossless´ frame work. Different from most of the existing visually lossless compression algorithm, the proposed algorithm can finally reach lossless stage thus suitable for medical application. The proposed algorithm exploits human visual sensitivity to different contrast. Based on this visual rather than mathematical significance, it adaptively determines the discriminating threshold for every wavelet coefficients. Compared with current bit-plane embedded coding algorithm, the proposed algorithm can get better image quality at given bit rate via arranging the bits of visible human apparatus before the bits depicting the invisible noises in dark background though the latter may be mathematically more significant than the former. From this point of view, it also achieves the goal of transmitting the region of interest (ROI) first without any side information for this information can be obtained from the already received coefficients from LL subband.