【作者】 李晓梅；

【导师】 刘伯强；

【作者基本信息】 山东大学 ， 生物医学工程， 2009， 硕士

【摘要】 随着现代医院数字化进程的加快,数字化医学图像在医院的诊断和治疗中发挥着越来越重要的作用。数字化医学图像的广泛应用,产生了大量的数据,给图像的存储和传输技术提出了严峻的挑战。因此,研究有针对性的医学图像压缩技术具有重要的实际意义。根据医学图像的特点和医学图像对压缩编码方案的特殊要求,本文基于小波变换,嵌入式图像编码,图像分割和感兴趣区域编码的各种理论方法,以从山东省立医院采集到的CT图像,MR图像和超声图像为主要研究对象,结合临床的实际需求,重点解决医学图像高图像质量和高压缩比之间的矛盾,并通过matlab编程对研究过程中涉及的算法进行了实现和分析。论文的主要工作如下:(1)研究整数小波变换在医学图像压缩中的应用。由于传统的小波变换需要进行小波系数的舍入运算,无法重建原始无损的图像,本文通过对整数小波变换理论和医学图像特征的分析研究,采用既能实现无损压缩又能实现有损压缩的5/3双正交小波基,通过小波分解后各子带的信息熵值和计算量来综合确定小波变换的级数,采用周期对称延拓方式实现了提升方案下的三级整数小波变换,使能量集中到少数小波系数上。实验结果表明利用整数小波变换进行图像压缩可以避免计算机的舍入误差,有利于医学图像的无损压缩。通过实验方法分析了图像小波系数的分布特点,为选择合适的压缩编码方案奠定了基础。(2)研究嵌入式医学图像压缩编码算法。通过对图像小波变换后小波系数分布特点的分析,本文发现它非常适合于嵌入式图像编码算法。在深入分析嵌入式零树小波编码算法和多级树集合分裂(SPIHT)编码算法优缺点的基础上,将SPIHT编码和自适应算术编码相结合,提出了基于整数小波变换的静态医学图像压缩编码方案,并利用此方案对四种不同类型的医学图像(细胞图像,CT图像,MR图像和超声图像)进行了压缩实验,给出了与其它编码方法的比较结果,仿真结果验证了该编码方案对医学图像压缩的有效性。(3)研究基于感兴趣区域的医学图像压缩编码算法。对于医学图像,医生往往仅对病变区域感兴趣,为了更好的解决医学图像高图像质量和高压缩比之间的矛盾,本文在提出的静态医学图像压缩编码方案中引进感兴趣区域编码,设计了更具实用性和兼容性的医学图像感兴趣区域编码方案。将一幅医学图像划分为感兴趣区域(病变区域)和背景区域两部分,针对JPEG 2000标准中一般位移法和最大位移法的优缺点,本文提出将图像分割与图像压缩相结合,采用主动轮廓模型算法提取病变区域作为感兴趣区域,并用回溯法生成掩模,解决医学图像压缩技术中感兴趣区域自动提取的问题。采用最大位移法进行了压缩实验,并对实验结果进行了分析。另外,考虑到医学图像周围的相关组织和背景也为正确诊断提供了参照,将能够灵活调整感兴趣区域和背景区域重要性的通用的部分重要位平面偏移方法应用到医学图像压缩中,并将实验结果与最大位移法进行了比较。实验结果表明,利用此方法既可以保证诊断信息有良好的视觉效果,同时又克服了最大位移法的缺点,在提高压缩率的基础上,有效实现了对感兴趣区域和背景区域压缩质量的灵活调整,为医学图像压缩方法提供了一种新思路。论文最后对全文进行了总结,并对下一步的研究工作进行了展望。更多还原

【Abstract】 With the acceleration of digital process in modern hospital, digital medical images are playing an increasingly important role in the diagnostic and treatment of hospital. Large amount of data are produced in the widespread application of medical image and they bring tough challenges to the storage and transmission of medical image. So it is of great practical significance to research targeted medical image compression techniques.According to the characteristics of medical image and the special needs of medical image compression scheme, based on theories and methods of wavelet transform, embedded image coding, image segmentation and region of interest coding, CT, MR and ultrasound images obtained from Shandong Provincial Hospital were mainly studied in this paper. Combining with clinical requirement, this paper mainly solved contradiction between high image quality and high compression ratio of medical image and the algorithms involved in research process were realized and analyzed by matlab programming. The main contents are listed below:(1) Research on the application of integer wavelet transform in medical image compression. Because traditional wavelet transform need to make rounding operation for wavelet coefficients, it can not reconstruct original lossless images. By analyzing integer wavelet transform theories and medical image characteristics, 5/3 biorthogonal wavelet basis, which can realize either lossless compression or lossy compression, is adopted in this paper. According to the information entropy value of each sub-band and computational complexity, the level of wavelet decomposition is comprehensively determined. Symmetric periodic extension is chose to realize three level integer wavelet transform based on lifting scheme and image energy concentrate on few wavelet coefficients. Experimental results show that the application of integer wavelet transform in image compression can avoid rounding error of computer and it is beneficial to medical image lossless compression. The distribution of wavelet coefficients lays the foundation for suitable coding scheme.(2)Research on coding algorithms of embedded medical image. According to the distribution characteristics of wavelet coefficients, it is very suitable for embedded image coding algorithm. On the basis of thorough analysis of advantage and disadvantage for embedded zero tree wavelet coding and set partition in hierachical trees coding, by combing SPIHT coding and adaptive arithmetic coding, a still medical image coding scheme based on integer wavelet transform is proposed in this paper. The compressive experiment to four different types of medical image (cell image, CT image, MR image and ultrasound image) is conducted and comparison results with other coding methods show that the scheme is very effective for medical image compression.(3)Research on medical image coding algorithms based on region of interest. Doctors are always only interested in lesion areas of medical image. In order to solve contradiction between high image quality and high compression ratio of medical image better, region of interest coding is introduced in the proposed still medical image coding scheme and ROI coding scheme is designed. Medical image is divided into region of interest area (lesion area) and background area. According to the advantage and disadvantage of Maxshift method and general scaling based method in JPEG 2000 standard, the method of combing image segmentation and image compression is presented in this paper. Active contour model is used to extract ROI region, backtracking is adopted to generate mask to solve the problem of ROI automatic extraction in medical image compression techniques. Maxshift method is adopted and experimental results are analyzed. In addition, considering that related tissue and background of medical image also provide reference for accurate diagnosis, Generalized Partial Bitplanes Shift Method is applied to medical image compression, which can adjust the significance of ROI and background region flexibly. Compared with Maxshift method, experimental results show that this method can guarantee the perfect visual effect of diagnostic information and can overcome the disadvantage of Maxshift. On the basis of improving compression rate, it can realize the flexible adjustment of image quality between ROI area and background area effectively and provide a new method for medical image compression. Finally, a conclusion of the main contents is made and the future research directions in this field are proposed.更多还原