Vela Res., St. Petersburgh, FL.
IEEE Trans Med Imaging. 1996;15(4):555-9. doi: 10.1109/42.511758.
Lossless compression of medical images using a proposed differentiation technique is explored. This scheme is based on computing weighted differences between neighboring pixel values. The performance of the proposed approach, for the lossless compression of magnetic resonance (MR) images and ultrasonic images, is evaluated and compared with the lossless linear predictor and the lossless Joint Photographic Experts Group (JPEG) standard. The residue sequence of these techniques is coded using arithmetic coding. The proposed scheme yields compression measures, in terms of bits per pixel, that are comparable with or lower than those obtained using the linear predictor and the lossless JPEG standard, respectively, with 8-b medical images. The advantages of the differentiation technique presented here over the linear predictor are: 1) the coefficients of the differentiator are known by the encoder and the decoder, which eliminates the need to compute or encode these coefficients, and 21 the computational complexity is greatly reduced. These advantages are particularly attractive in real time processing for compressing and decompressing medical images.
利用提出的差分技术对医学图像进行无损压缩的研究。该方案是基于计算相邻像素值之间的加权差。评估了所提出的方法对磁共振（MR）图像和超声图像进行无损压缩的性能，并将其与无损线性预测器和无损联合图像专家组（JPEG）标准进行了比较。这些技术的残差序列使用算术编码进行编码。对于 8 位医学图像，所提出的方案在每像素位数方面的压缩度量值可与线性预测器或无损 JPEG 标准获得的压缩度量值相媲美或更低。与线性预测器相比，这里提出的差分技术的优势在于：1） 微分器的系数为编码器和解码器所熟知，这消除了计算或编码这些系数的需要，并且 21 计算复杂度大大降低。这些优势在用于压缩和解压缩医学图像的实时处理中特别有吸引力。
