字典学习中字典尺度对DICOM图像压缩的影响

随着医院数字化医疗进程的加快,医学影像的数据量日益增大,医学影像资料的存储空间和获取速度受到很大的限制。文章在研究主流字典学习算法基础上,提出使用不同尺度的MOD、K-SVD、ILS-DLA、RLS-DLA字典算法对DICOM图像进行压缩存储,以及恢复再现的方法。与经典的JPEG和JPEG2000压缩算法相比,字典学习算法压缩和恢复效果较好,特别是采用较小尺度的字典时,压缩效果更为突出:当压缩比为20时,采用4×4尺度的RLS-DLA字典,论文算法的峰值信噪比(PSNR)较JPEG算法高出7.8dB,比JPEG2000算法高出1dB。

Effects of dictionary scale on dictionary learning for DICOM image compression

With the accelerated developing of hospital digital medical, the amount of medical imaging data grows dramatically, which affects the data storage space and access speed. This paper proposes a new design which uses different scales dictionaries of MOD, K-SVD, ILS-DLA, RLS-DLA for digital imaging and communications in medicine (DICOM) image compression storage and restore methods based on dictionary learning. Compared with the traditional algorithms JPEG and JPEG2000, the proposed method has better performance, especially when the dictionary scale is smaller. For example, when the compression ratio is 20, using 4×4 dictionary scale, the peak signal to noise ratio (PSNR) of the proposed method is 7.8 dB higher than that of JPEG, and 1dB than JPEG2000.