计算机工程

• 图形图像处理 • 上一篇    下一篇

基于快速稀疏表示的医学图像压缩

赵海峰1,2,鲁毓苗1,陆 明1,陈思宝1,2   

  1. (1. 安徽大学计算机科学与技术学院,合肥 230601;2. 安徽省工业图像处理与分析重点实验室,合肥 230039)
  • 收稿日期:2013-06-28 出版日期:2014-04-15 发布日期:2014-04-14
  • 作者简介:赵海峰(1972-),男,副教授、博士,主研方向:医学图像处理,模式识别;鲁毓苗、陆 明,硕士研究生;陈思宝, 副教授。
  • 基金项目:
    国家自然科学基金资助项目(61272152, 61202228, 61300057);安徽省自然科学基金资助项目(1208085MF109);2013年留学人员科技活动择优基金资助项目;安徽省高校自然科学研究基金资助重点项目(KJ2013A007)。

Medical Image Compression Based on Fast Sparse Representation

ZHAO Hai-feng  1,2, LU Yu-miao  1, LU Ming  1, CHEN Si-bao  1,2   

  1. (1. School of Computer Science and Technology, Anhui University, Hefei 230601, China; 2. Key Lab of Industrial Image Processing & Analysis of Anhui Province, Hefei 230039, China)
  • Received:2013-06-28 Online:2014-04-15 Published:2014-04-14

摘要: 随着数字医学图像数据量的日益增大,有必要采取一定的图像压缩技术进行压缩存储。为此,提出基于快速稀疏表示的医学图像压缩方法。使用K-奇异值分解算法构造医学图像过完备字典,采用批量正交匹配追踪(Batch-OMP)算法进行稀疏编码。该方法只需要存储稀疏编码非零位置的系数信息,利用过完备字典即可实现原始医学图像的重构。实验结果表明,该方法可提高图像稀疏编码的速度,与正交匹配追踪(OMP)算法相比可提速40%左右,并且图像重构效果优于联合图像专家组(JPEG)算法和多级树集合分裂(SPIHT)算法的压缩效果,相对JPEG压缩的图像峰值信噪比平均提高18%,相对SPIHT算法平均提高50%。

关键词: 稀疏表示, 医学图像压缩, K-SVD算法, 稀疏编码, OMP 算法, Batch-OMP算法

Abstract: With the increasing growth of the digital medical image data, more image processing technology is needed to implement compressive storage. However, current image compression methods do not consider the characteristics of medical image. Aiming at this problem, a method of medical image compression based on fast sparse representation is proposed. It uses the K-Singular Value Decomposition(K-SVD) algorithm to construct an over-complete dictionary for sparse representation, and uses the Batch Orthogonal Matching Pursuit(Batch-OMP) algorithm for sparse coding. Only the coefficients information in the nonzero position of sparse coding is needed to be stored for recovering the original medical images perfectly with the over-complete dictionary. Experimental results show that the proposed method can speed up about 40% compared with Orthogonal Matching Pursuit(OMP) when performing image compression. Furthermore, the results of image reconstruction show that the proposed method increases the Peak Signal to Noise Ratio(PSNR) of the compressed images by an average of 18% and 50% compared with Joint Photographic Experts Group(JPEG) algorithm and Set Partitioning In Hierarchical Trees(SPIHT) algorithm respectively, indicating that the proposed method has a better performance than JPEG and SPIHT.

Key words: sparse representation, medical image compression, K-SVD algorithm, sparse coding, OMP algorithm, Batch-OMP algorithm

中图分类号: