基于Curvelet变换和SPIHT算法的医学图像感兴趣区压缩

摘要

Purpose To propose a novel compression method for region of interest (ROI) based on Curvelet transform and SPIHT algorithm. Materials and Methods The ROI was firstly extracted without compression, and Curvelet transform was applied for the background regions. The Curvelet coefifcients were coded using SPIHT algorithm. Then the images after compression are obtained by inverse Curvelet transform. The ROI and the background were ifnally overlapped to get the full compressed image. Effect of ROI compression and overall compression were compared, as well as the Curvelet transform and wavelet transform, based on peak signal noise ratio. Results The ROI compression highlighted the region of interest and the visual effect was superior to the overall compression. The peak signal to noise of Curvelet transform was higher than that of wavelet transforms, and the compressed images were more clear for the same proportion. Conclusion ROI compression based on Curvelet transform and SPIHT algorithm can achieve efficient compression images without losing important diagnostic information, which complies with the requirement of high precision and high quality of medical image compression.%目的：提出基于Curvelet变换和多级树集合分裂排序（SPIHT）算法的图像感兴趣区（ROI）压缩方法，并应用于医学图像压缩。资料与方法算法流程首先对图像ROI进行提取，保留ROI不压缩，对背景区域进行Curvelet变换，采用SPIHT算法对Curvelet系数进行编码；然后进行Curvelet逆变换得到有损压缩后的图像；最后将ROI区域与背景区域叠加，得到压缩后的完整图像。采用峰值信噪比作为评价指标，比较ROI压缩和整体压缩的效果，以及小波变换和Curvelet变换用于图像压缩的效果差异。结果分别对测试图像和医学图像的压缩结果进行比较，采用ROI压缩的视觉效果优于整体压缩的效果，更能突出ROI；而采用Curvelet变换压缩的峰值信噪比高于小波变换压缩，相同比例的压缩图像也更清晰。结论基于Curvelet变换和SPIHT算法的ROI压缩可在保证不丢失重要诊断信息的前提下实现图像的高效压缩，符合医学图像压缩的高精度、高质量要求。