基于医学透视图像的股骨上段特征提取
|
摘要
医学图像处理是计算机视觉领域的一个重要分支,是数字图像技术在生物医学工程中的应用。对医学图像进行特征提取,可以有效地指导临床诊断,同时也为人体信息数字化的科学研究打下基础。
不同的医学股骨透视图像特点不同,故采用不同的方法提取其特征。针对X光图像不够清晰、不易对股骨进行直接分割的特点,采用数据拟合的方法,结合股骨特征模型直接从图像数据中拟合出股骨的特征参数。针对图像比较清晰的序列CT图像,采用先对二维股骨图像进行轮廓分割,再对多层轮廓结果进行特征提取的方法。
在提取股骨单帧轮廓时,提出自适应辐射线法分割类圆形股骨轮廓。根据图像拍摄初试条件确定骨骼大致位置,利用少量辐射线对类圆形股骨进行精确定位并计算得到类圆形的均半径,使用数量适当的辐射线扫描确定股骨边缘点,再结合边缘点可信度及圆滑指数对边缘点进行修正。最终实现了对类圆形股骨轮廓的准确提取。
在提取股骨序列轮廓时提出关联帧间跟踪法,结合相邻帧的股骨轮廓确定当前帧股骨轮廓。根据相邻帧的股骨轮廓边界矩形,确定当前帧轮廓包围矩形；依据相邻帧股骨轮廓将当前帧包围矩形内图像转化为边缘极展开图像；结合边缘点的可信度以及圆滑指数等对边缘点进行修正确定股骨轮廓；对双向股骨轮廓信息进行融合,确定最终股骨轮廓,并据此轮廓线实现骨骼的三维重建,获得完整的三维骨骼信息。
最后,对所做的研究工作进行了总结,提出了需要开展的后续工作的思路,对进一步的研究具有一定的指导意义。
As one research branch in the field of computer vision, processing medical image is an important application of computer graphics and digital image processing in the biomedical engineering. Feature extraction, the foundation of digital research, can be used to guide Clinical Diagnosis effectively.
Medical perspective images having different features result in different methods used to extract the skeletal contour for femurs. Data fitting, combining with feature models of femur, is applied to acquire the feature parameters of cloudy X-Ray images. While, for distinct CT sequence images, extracting skeletal contour for femurs in 2-D images, followed by feature extraction of multilayer contour is a better choice.
The method of radiating lines with self-adapting is used to acquire the contour of femur similar to a circle object when extracting contour of femur in a single image. Firstly, according to initial conditions can be used to fix the position of a femur approximately. Furthermore a few radiating lines go a step further to get more accurate the position of the femur and gain the radius of the contour liking a circle. Finally, reasonable quantity of radiating lines scanning is used to confirm the edge points of the femur and the results will be revised on the basis of the reliability of edge points and smoothing index. Therefore, the contour of a femur liking a circle can be obtained.
The current contour of a femur can be gained by tracking a sequence of frames and the contour of a femur in consecutive frames in the process of extracting contours in sequence images. Rectangular region of a femur in current frame can be confirmed by the rectangular regions of a femur in consecutive frames, and then the edges of the image based on its polar coordinates will be acquired. These edge points extracted will be revised in further according to the reliability and smoothing index. After mixing the contours of bidirectional femur together, the ultimate contour of a femur will be gained and these results will be used to realize the 3-D reconstruction of a femur and conclude the whole 3-D femur information.
Finally, on the basis of the summarized work, the essay has put forward some useful suggestions in the later study work, which is in great favor of guiding the further study work.
不同的医学股骨透视图像特点不同,故采用不同的方法提取其特征。针对X光图像不够清晰、不易对股骨进行直接分割的特点,采用数据拟合的方法,结合股骨特征模型直接从图像数据中拟合出股骨的特征参数。针对图像比较清晰的序列CT图像,采用先对二维股骨图像进行轮廓分割,再对多层轮廓结果进行特征提取的方法。
在提取股骨单帧轮廓时,提出自适应辐射线法分割类圆形股骨轮廓。根据图像拍摄初试条件确定骨骼大致位置,利用少量辐射线对类圆形股骨进行精确定位并计算得到类圆形的均半径,使用数量适当的辐射线扫描确定股骨边缘点,再结合边缘点可信度及圆滑指数对边缘点进行修正。最终实现了对类圆形股骨轮廓的准确提取。
在提取股骨序列轮廓时提出关联帧间跟踪法,结合相邻帧的股骨轮廓确定当前帧股骨轮廓。根据相邻帧的股骨轮廓边界矩形,确定当前帧轮廓包围矩形；依据相邻帧股骨轮廓将当前帧包围矩形内图像转化为边缘极展开图像；结合边缘点的可信度以及圆滑指数等对边缘点进行修正确定股骨轮廓；对双向股骨轮廓信息进行融合,确定最终股骨轮廓,并据此轮廓线实现骨骼的三维重建,获得完整的三维骨骼信息。
最后,对所做的研究工作进行了总结,提出了需要开展的后续工作的思路,对进一步的研究具有一定的指导意义。
As one research branch in the field of computer vision, processing medical image is an important application of computer graphics and digital image processing in the biomedical engineering. Feature extraction, the foundation of digital research, can be used to guide Clinical Diagnosis effectively.
Medical perspective images having different features result in different methods used to extract the skeletal contour for femurs. Data fitting, combining with feature models of femur, is applied to acquire the feature parameters of cloudy X-Ray images. While, for distinct CT sequence images, extracting skeletal contour for femurs in 2-D images, followed by feature extraction of multilayer contour is a better choice.
The method of radiating lines with self-adapting is used to acquire the contour of femur similar to a circle object when extracting contour of femur in a single image. Firstly, according to initial conditions can be used to fix the position of a femur approximately. Furthermore a few radiating lines go a step further to get more accurate the position of the femur and gain the radius of the contour liking a circle. Finally, reasonable quantity of radiating lines scanning is used to confirm the edge points of the femur and the results will be revised on the basis of the reliability of edge points and smoothing index. Therefore, the contour of a femur liking a circle can be obtained.
The current contour of a femur can be gained by tracking a sequence of frames and the contour of a femur in consecutive frames in the process of extracting contours in sequence images. Rectangular region of a femur in current frame can be confirmed by the rectangular regions of a femur in consecutive frames, and then the edges of the image based on its polar coordinates will be acquired. These edge points extracted will be revised in further according to the reliability and smoothing index. After mixing the contours of bidirectional femur together, the ultimate contour of a femur will be gained and these results will be used to realize the 3-D reconstruction of a femur and conclude the whole 3-D femur information.
Finally, on the basis of the summarized work, the essay has put forward some useful suggestions in the later study work, which is in great favor of guiding the further study work.
引文
[1]罗述谦,周果宏.医学图象处理与分析[M].北京：科学出版社.2003：1-3.
[2]张钟元,柴伟明.螺旋CT三维重建在复杂骨折诊疗中的价值[J].上海第二医科大学学报,2004,12(24)：1047-104.
[3]邱甜甜,黄远亮.牙CT在口腔临床诊治中的应用[J].口腔颌面外科杂志,2008,10(18)：373-376.
[4]张燕中,李俊峰.膝关节损伤的MRI诊断分析[J].长治医学院学报,2008,12(22)：454-456.
[5]李伟,陈武凡.基于凸轮的交互式脑部肿瘤MRI自动三维分割[J].南方医科大学学报,2009,29(1)：140-143.
[6]潘中允.PET诊断学[M].北京：人民卫生出版社,2005：35-38.
[7]罗述谦.x射线成像技术在医学中应用[J].物理.2007,8(39)：602-608.
[8]宋钢,邓小玖.浅谈CT[J].现代物理知识.2008,02(17)：20-22.
[9]吕厚山.人工关节外科学[M].北京：科学出版社,2001：1-414.
[10]座谈纪要,髋关节人工置换[J].中华外科杂志.2000,7(4)：8-14.
[11]钱齐荣等,新型无柄人工髋关节生物固定的实验研究[J].第二军医大学学报,2000,21(7)：8-14.
[12]David, Hungerford, Lynnec Jones. The relational for cemented total replacement[J]. Orthopedics Clinics of North America,1993,4:75～79.
[13]王纪湘,符广敏,王以进.介绍一种新型人工髋关节—螺旋髋[C].生物摩擦学与人工关节学术研讨会,上海,2000,9：59-66,.
[14]Chun-HsiungShih New concepts-biomechanical studies of a newly designed femoral prosthesis[J]. Clinical Biomechanics 1997,12(8):482～490.
[15]杨加,吴祈耀.医学图形图像技术及图像分割[J].世界医疗器械.2000,6(4)：24-28.
[16]罗述谦,周果宏.医学图象处理与分析[M].北京：科学出版社：2003,65-66.
[17]Pikaz A, Averbuch A. Digital image thresholding based on topological stable state[J]. Pattern Recognition,1996,29(5):829-843.
[18]Corneloup G, Moysan J, Maynin I E. B-SCAN image segmentation by thresholding using cooccurrence matrix analysis[J]. Pattern Recognition,1996, 29(2):281～296.
[19]章毓晋.图像分割[M].北京：科学出版社,2001：48-50.
[20]Pham DL, Xu CY. A survey of current methods in medical image segmentation[J]. Annual Review of Biomedical Engineering.2000,2(7):315～319.
[21]贾允,丁艳,刘泽平.改进图像阈值分割算法的研究[J].光学技术,2005,31(1)：155-157.
[22]Kim D Y, Kim J H, Noh S M, etc. Pulmonary nodule detection using chest CT images[J]. Act Radio,2003,44(3):252～257.
[23]陆剑锋,林海,潘志庚,等.自适应区域生长算法在医学图像分割中的应用[J].计算机辅助设计与图形学学报,2005,17(10)：2168-2173.
[24]Fan L, Braden G A, Herrington D M. Nonlinear wavelet filter for intracoronary ultrasound images[C]. Computers in Cardiology, Indianapolis,1996:41-44.
[25]朱志松,朱龙彪,黄希.基于数学形态变换的骨骼CT图像边缘提取[J].微计算机信息,2006,22(10-3)：298-300.
[26]M Kass, A Witkin, D Terzopoulos. Snake:Active Contour Models[J]. International Journal of Computer Vision,1988,1(4):321～331
[27]L D Cohen. An Active Contour Models and Balloons[J]. CVGIP Image Understanding,1991,53(2):211～218.
[28]Lobregt S, Viergever MA. A Discrete Dynamic Contour Model[J]. IEEE Trans. on Medical images,1995,14(1):12～24.
[29]Xu C, Prince J L. Generalized gradient vector flow external forces for active contours[J]. Signal Processing,1998,71:131～139.
[30]Jierong Cheng, Say Wei Foo. Dynamic Directional Gradient Vector Flow for Snake[J]. IEEE Transactions on Image Processing,2006,15(6):1563～1571.
[31]Caselles Vicentl, Kimmel Ronl, Sapiro Guillermol. Geodesic Active Contours[J]. IEEE International Conference on Computer Vision,1995,6: 694～699.
[32]张治国,周越,谢凯.一种基于Muford-Shah模型的脑肿瘤水平集分割算法[J].上海交通大学学报,2005,39(12)：1955-1958.
[33]聂生东,聂斌.磁共振脑图像模糊聚类分割中的参数选择问题研究[J].中国医学物理学杂志,2004,21(2)：75-82.
[34]Shiying Hu, Eric A. Hoffman, and Joseph M Reinhardt. Automatic Lung Segmentation for Accurate Quantitation of Volumetric X-Ray CT Images[J]. IEEE Engineering in Medicine and Biology Society,2001,20(6):490～498.
[35]Bae KT, Giger ML, Chen CT, et al. Automatic segmentation of liver structure in CT images[J]. Medical Physics,1993,20(1):71～75.
[36]Herment A, Mousseaus E, Dumee P, et al. Automatic detection of left ventricular borders on electron beam CT sequential cardiac images using an adaptive algorithm[J]. Computerized Medical Imaging and Graphics,1998,22(4): 291～294.
[37]Liu S., Ma W. Seed-growing segmentation of 3-D surfaces from CT-contour data[J]. Computer Aided Design,1999,31(8):517～519.
[38]Cagnoni S, Dobrzeniecki AB, PoliR, et al. Genetic algorithm-based interactive segmentation of 3D medical images[J]. Image and Vision Computing,1999, 17(12):881～885.
[39]江少峰,陈武凡,冯前进,等.序列颅脑CT图像颅腔内结构自动化提取及分割[J].南方医科大学学报,2007,27(12)：1805-1808.
[40]张飞,樊瑜波,蒲放,等.牙颌CT图像序列中牙的半自动分割方法[J].生物医学杂志,2007,24(1)：15-18.
[41]吴良武,侯健华,张勇,等.使用邻域运算从CT图像中分割骨骼[J].中国生物医学工程报,2003,20(3)：199-202.
[42]Zhi-Qiang Liu, Hui Lee Liew, John G Clement, et al. Bone Image Segmentation[J]. IEEE transactions on biomedical engineering 1999,46(5): 565～573.
[43]Fiebich M, Straus C M, Sehgal V, et al. Automatic bone segmentation technique for CT angiographic studies[J]. Journal of Computer Assisted Tomography,1999, 23(1):155～161.
[44]Sebastian T B, Tek H, Crisco J J, et al. Segmentation of carpal bones from 3D CT images using skeletally coupled deformable models[C]. Medical Image Computing and Computer-Assisted Intervention-MICCAI'98, Cambridge:1998, 1184～1194.
[45]Michael D J, Nelson A C. a model-based system for automatic segmentation of bones from digital hand radiographs[J]. IEEE Transactions on Medical Imaging, 1989,8(1):64～69.
[46]Baniasadipour A, Zoroofi R A, Sato Y, et al. A fully automated method for segmentation and thickness map estimation of femoral and acetabular cartilages in 3D CT images of the hip[C].2007 5th International Symposium on Image and Signal Processing and Analysis, Istanbul:2007,244-249.
[47]陆莎,张欣.基于数字图像的足部骨骼建模研究[J].西安工程大学学报, 2008,4(22-2)：187-191.
[48]Weiwei Song, Guanhua Li, Zongying Ou, et al. Model-Based Segmentation of Femoral Head and Acetabulum from CT Images[C]. International Conference on Complex Medical Engineering, Beijing:2007,586～590.
[49]宋卫卫,苏铁明,欧宗瑛,等.基于CT图像反求技术的人体股骨头修复建模[J].大连理工大学学报,2009,49(2)：222-227.
[50]刘斌,宋卫卫,欧宗瑛,等.基于提出拟合股骨头修复建模[J].东南大学学报,2008,38(1)：58-63.
[51]韦轶群,杨杰.基于自适应Graph Cuts的自动股骨头分割[J].上海交通大学学报,2009,43(3)：465-469.
[52]何青,刘允才.股骨模型的三维重建与变形[J].计算机工程,2007,33(13)：221-223.
[53]孙海波,尚仕波,姚新港.CAx技术在人体股骨逆向工程中的应用[J].工程图学学报,2008,3：12-16.
[54]叶哲伟,杨述华,周均清,等.复杂髋关节骨折CT图像的三维重建[J].中国组织工程研究与临床康复,2008,17(12)：3282-3284.
[55]凌玲,王成焘.股骨两维有限元的x光片装正处理[J].机械设计与研究,1996,2：34-37.
[56]马如宇.人体髋关节系统定制型假体设计理论与方法[D].上海：上海交通大学,2005.
[57]董天华.髋关节外科学[M].郑州：郑州大学出版社,2005：65-75.
[58]李浴中.X线摄影感光化学及显像过程[M].北京：人民军医出版社,1992：33-42.
[59]孟悛非.医学影像学[M].北京：高等教育出版社,2004：28-33.
[60]罗三定,肖飞.不规则类圆形团块目标图像识别的新方法[J].中南大学学报,2004,35(4)：632-637.
[61]张会章.基于视觉感知的图像理解方法研究[D].西安：西北工业大学,2003.
[2]张钟元,柴伟明.螺旋CT三维重建在复杂骨折诊疗中的价值[J].上海第二医科大学学报,2004,12(24)：1047-104.
[3]邱甜甜,黄远亮.牙CT在口腔临床诊治中的应用[J].口腔颌面外科杂志,2008,10(18)：373-376.
[4]张燕中,李俊峰.膝关节损伤的MRI诊断分析[J].长治医学院学报,2008,12(22)：454-456.
[5]李伟,陈武凡.基于凸轮的交互式脑部肿瘤MRI自动三维分割[J].南方医科大学学报,2009,29(1)：140-143.
[6]潘中允.PET诊断学[M].北京：人民卫生出版社,2005：35-38.
[7]罗述谦.x射线成像技术在医学中应用[J].物理.2007,8(39)：602-608.
[8]宋钢,邓小玖.浅谈CT[J].现代物理知识.2008,02(17)：20-22.
[9]吕厚山.人工关节外科学[M].北京：科学出版社,2001：1-414.
[10]座谈纪要,髋关节人工置换[J].中华外科杂志.2000,7(4)：8-14.
[11]钱齐荣等,新型无柄人工髋关节生物固定的实验研究[J].第二军医大学学报,2000,21(7)：8-14.
[12]David, Hungerford, Lynnec Jones. The relational for cemented total replacement[J]. Orthopedics Clinics of North America,1993,4:75～79.
[13]王纪湘,符广敏,王以进.介绍一种新型人工髋关节—螺旋髋[C].生物摩擦学与人工关节学术研讨会,上海,2000,9：59-66,.
[14]Chun-HsiungShih New concepts-biomechanical studies of a newly designed femoral prosthesis[J]. Clinical Biomechanics 1997,12(8):482～490.
[15]杨加,吴祈耀.医学图形图像技术及图像分割[J].世界医疗器械.2000,6(4)：24-28.
[16]罗述谦,周果宏.医学图象处理与分析[M].北京：科学出版社：2003,65-66.
[17]Pikaz A, Averbuch A. Digital image thresholding based on topological stable state[J]. Pattern Recognition,1996,29(5):829-843.
[18]Corneloup G, Moysan J, Maynin I E. B-SCAN image segmentation by thresholding using cooccurrence matrix analysis[J]. Pattern Recognition,1996, 29(2):281～296.
[19]章毓晋.图像分割[M].北京：科学出版社,2001：48-50.
[20]Pham DL, Xu CY. A survey of current methods in medical image segmentation[J]. Annual Review of Biomedical Engineering.2000,2(7):315～319.
[21]贾允,丁艳,刘泽平.改进图像阈值分割算法的研究[J].光学技术,2005,31(1)：155-157.
[22]Kim D Y, Kim J H, Noh S M, etc. Pulmonary nodule detection using chest CT images[J]. Act Radio,2003,44(3):252～257.
[23]陆剑锋,林海,潘志庚,等.自适应区域生长算法在医学图像分割中的应用[J].计算机辅助设计与图形学学报,2005,17(10)：2168-2173.
[24]Fan L, Braden G A, Herrington D M. Nonlinear wavelet filter for intracoronary ultrasound images[C]. Computers in Cardiology, Indianapolis,1996:41-44.
[25]朱志松,朱龙彪,黄希.基于数学形态变换的骨骼CT图像边缘提取[J].微计算机信息,2006,22(10-3)：298-300.
[26]M Kass, A Witkin, D Terzopoulos. Snake:Active Contour Models[J]. International Journal of Computer Vision,1988,1(4):321～331
[27]L D Cohen. An Active Contour Models and Balloons[J]. CVGIP Image Understanding,1991,53(2):211～218.
[28]Lobregt S, Viergever MA. A Discrete Dynamic Contour Model[J]. IEEE Trans. on Medical images,1995,14(1):12～24.
[29]Xu C, Prince J L. Generalized gradient vector flow external forces for active contours[J]. Signal Processing,1998,71:131～139.
[30]Jierong Cheng, Say Wei Foo. Dynamic Directional Gradient Vector Flow for Snake[J]. IEEE Transactions on Image Processing,2006,15(6):1563～1571.
[31]Caselles Vicentl, Kimmel Ronl, Sapiro Guillermol. Geodesic Active Contours[J]. IEEE International Conference on Computer Vision,1995,6: 694～699.
[32]张治国,周越,谢凯.一种基于Muford-Shah模型的脑肿瘤水平集分割算法[J].上海交通大学学报,2005,39(12)：1955-1958.
[33]聂生东,聂斌.磁共振脑图像模糊聚类分割中的参数选择问题研究[J].中国医学物理学杂志,2004,21(2)：75-82.
[34]Shiying Hu, Eric A. Hoffman, and Joseph M Reinhardt. Automatic Lung Segmentation for Accurate Quantitation of Volumetric X-Ray CT Images[J]. IEEE Engineering in Medicine and Biology Society,2001,20(6):490～498.
[35]Bae KT, Giger ML, Chen CT, et al. Automatic segmentation of liver structure in CT images[J]. Medical Physics,1993,20(1):71～75.
[36]Herment A, Mousseaus E, Dumee P, et al. Automatic detection of left ventricular borders on electron beam CT sequential cardiac images using an adaptive algorithm[J]. Computerized Medical Imaging and Graphics,1998,22(4): 291～294.
[37]Liu S., Ma W. Seed-growing segmentation of 3-D surfaces from CT-contour data[J]. Computer Aided Design,1999,31(8):517～519.
[38]Cagnoni S, Dobrzeniecki AB, PoliR, et al. Genetic algorithm-based interactive segmentation of 3D medical images[J]. Image and Vision Computing,1999, 17(12):881～885.
[39]江少峰,陈武凡,冯前进,等.序列颅脑CT图像颅腔内结构自动化提取及分割[J].南方医科大学学报,2007,27(12)：1805-1808.
[40]张飞,樊瑜波,蒲放,等.牙颌CT图像序列中牙的半自动分割方法[J].生物医学杂志,2007,24(1)：15-18.
[41]吴良武,侯健华,张勇,等.使用邻域运算从CT图像中分割骨骼[J].中国生物医学工程报,2003,20(3)：199-202.
[42]Zhi-Qiang Liu, Hui Lee Liew, John G Clement, et al. Bone Image Segmentation[J]. IEEE transactions on biomedical engineering 1999,46(5): 565～573.
[43]Fiebich M, Straus C M, Sehgal V, et al. Automatic bone segmentation technique for CT angiographic studies[J]. Journal of Computer Assisted Tomography,1999, 23(1):155～161.
[44]Sebastian T B, Tek H, Crisco J J, et al. Segmentation of carpal bones from 3D CT images using skeletally coupled deformable models[C]. Medical Image Computing and Computer-Assisted Intervention-MICCAI'98, Cambridge:1998, 1184～1194.
[45]Michael D J, Nelson A C. a model-based system for automatic segmentation of bones from digital hand radiographs[J]. IEEE Transactions on Medical Imaging, 1989,8(1):64～69.
[46]Baniasadipour A, Zoroofi R A, Sato Y, et al. A fully automated method for segmentation and thickness map estimation of femoral and acetabular cartilages in 3D CT images of the hip[C].2007 5th International Symposium on Image and Signal Processing and Analysis, Istanbul:2007,244-249.
[47]陆莎,张欣.基于数字图像的足部骨骼建模研究[J].西安工程大学学报, 2008,4(22-2)：187-191.
[48]Weiwei Song, Guanhua Li, Zongying Ou, et al. Model-Based Segmentation of Femoral Head and Acetabulum from CT Images[C]. International Conference on Complex Medical Engineering, Beijing:2007,586～590.
[49]宋卫卫,苏铁明,欧宗瑛,等.基于CT图像反求技术的人体股骨头修复建模[J].大连理工大学学报,2009,49(2)：222-227.
[50]刘斌,宋卫卫,欧宗瑛,等.基于提出拟合股骨头修复建模[J].东南大学学报,2008,38(1)：58-63.
[51]韦轶群,杨杰.基于自适应Graph Cuts的自动股骨头分割[J].上海交通大学学报,2009,43(3)：465-469.
[52]何青,刘允才.股骨模型的三维重建与变形[J].计算机工程,2007,33(13)：221-223.
[53]孙海波,尚仕波,姚新港.CAx技术在人体股骨逆向工程中的应用[J].工程图学学报,2008,3：12-16.
[54]叶哲伟,杨述华,周均清,等.复杂髋关节骨折CT图像的三维重建[J].中国组织工程研究与临床康复,2008,17(12)：3282-3284.
[55]凌玲,王成焘.股骨两维有限元的x光片装正处理[J].机械设计与研究,1996,2：34-37.
[56]马如宇.人体髋关节系统定制型假体设计理论与方法[D].上海：上海交通大学,2005.
[57]董天华.髋关节外科学[M].郑州：郑州大学出版社,2005：65-75.
[58]李浴中.X线摄影感光化学及显像过程[M].北京：人民军医出版社,1992：33-42.
[59]孟悛非.医学影像学[M].北京：高等教育出版社,2004：28-33.
[60]罗三定,肖飞.不规则类圆形团块目标图像识别的新方法[J].中南大学学报,2004,35(4)：632-637.
[61]张会章.基于视觉感知的图像理解方法研究[D].西安：西北工业大学,2003.