非荧光造影图像的高血压病灶提取方法
|
摘要
眼底一直是临床医学密切关注的区域,不少重要的全身疾病可在眼底引起特定的反应或并发症,临床上往往可以从眼底的某些特殊体征协助其他学科做出正确的诊断以及预后的估计。眼底荧光血管造影是眼科临床常用检查技术,但该方法造价较高并且在应用过程中各家都报道了一些副作用。因此,对非荧光造影图像的高血压病灶提取具有十分重要的意义。
本文采用了Visual C++系统,在对眼底图像的特点进行分析的基础上,做了大量的实验,对非荧光造影图像进行了图像增强、边缘检测、图像分割、去噪以及视盘定位等预处理,并实现了对非荧光造影眼底图像中的动静脉血管管径的测量,根据动静脉血管的管径比可以判断该人是否为高血压病患者,并对高血压病的视网膜病变期出现的白斑病灶进行比较准确的提取,在正常和异常之间做出明确鉴别,使对高血压病眼底的非荧光造影图像的研究标准化,从而大大加强眼底检查优势,方便了高血压患者,对高血压病的诊断及治疗具有重要的实践指导意义。
The development of image processing technology has been widely applied to the scientific research, industrial and agricultural production, military technology, government departments, medical hygiene and many other fields, further promoting the development of social productivity. Today, the multidisciplinary cross-integration of digital image processing make science and technology gradually infiltrate into other areas and it is inevitable to be applied by many other subjects. Both in theory and in practice, digital image processing has the great potential in information society. Medical image processing and analysis are always the focus and difficult problems for the research of image processing and analysis. By dint of the means of graph and image technology, the displaying method and quality of medical image got a tremendous improvement, which has been greatly improved the level of diagnosis. Optic fundus has been the focus of clinical medicine because retinal lesions are closely related to the diseases of other system of the whole body. Therefore, the special signs of optic fundus can be helpful to make a correct diagnosis and prognosis estimates.
The optic fundus is "barometer" which reflect the changes of hypertensive condition. According to observing the condition of retinal arterial vessel, such as the changes of blood vessel diameter and vascular wall, the phenomenon of cross and oppression, retinal hemorrhage, exudates and its property etc., people can acquaint the whole body vascular condition Fluorescence fundus angiography is commonly used on ophthalmic clinical examination, especially for pathologic diagnosis of optic fundus and the choice of treatment. However, some side effects were reported in theapplication of this technology. Fluorescence sodium which is used as the contrast agent will result in vary effects such as nausea, vomiting, numbness or shock, cardiac arrest and other adverse reactions. Therefore, it is significant to research the means of non-Fluorescence fundus angiography. Taking account of the accuracy, objectivity, reproducibility and universality of the results of optic fundus examination, it is vital to improve the analysis of fundus images especially the quantitative analysis for retinal vessels and the extraction of retinal exudates which is closely related to hypertension. Consequently, it is essential to take effective method to segregate optic fundus images and automatically analyze and measure vascular width. Based on the changes of retinal vascular pattern, this helps ophthalmologist to make early diagnosis, monitor treatment and estimate postoperative recovery with the great clinical significance.
In this paper, base on the analysis of the characteristics of the fundus image, a lot of experiments were finished with the Visual C + + system and extract two kinds of hypertension lesions were extracted from the non-fluorescence fundus angiography images.
(1) The measurements of the blood vessel diameter in the fundus images. The optic fundus images of the hypertensive sufferer are often very unclear because of complex blood vessels, retinal hemorrhage, exudates and such lesions represent. More noise also is interferential for the follow-up identification. Therefore, the fundus images must be pretreated to reduce unrelated information for the validity, veracity of exacting eigen value. In this paper, gray linear transform in green gray image was taken to enhance the definition in the retinal RGB image, Sobel operator was applied to detect images edge and the binary image processing was performed by the settledthreshold segmentation. But there are still a lot of noise points. Subsequently, noise was reduced to process the follow-up image. After that, image quality has been improved; image features has been enhanced; Noise has been reduced, the result of image processing and analyzing has been improved and operating speed has been accelerated. As the matching pattern of retinal vascular identified system, detecting the location of the retinal optic disc is the key to identification of the retinal vessels in the optic fundus images. The border of retinal optic disc has a simple geometric shape--round features. According to the geometric characteristics of the optic disc, the Hough transform was used to detect the optic disc and it is also helpful for the next step of the measurement of vascular diameter. To reduce the amount of computation, a rectangular region was designated as the working area according to the vascular distribution in the vicinity of the optic disc. In this paper, two methods of retinal vascular diameter were applied. The first method adopted by the least square method of vascular wall for curve fitting, choose center of the optic disc which the Hough transform detect as the center to draw a circle which intersect with the vascular wall. With this way, nodal striking distance is the vessel diameter. The second method adopted the closing operation of mathematical morphology for de-noising retinal images, to fill vein first without changes in the diameter of the artery and then to measure the diameter. According to the ratio of the average artery and vein diameter, the retinal image can be judged whether it is of Hypertension retinal images. By contrast, the first method is more applicable to the detection of poor continuity of the vascular wall; the second method is adaptive for anti-jamming.
(2) The lesions extraction of retinal exudates. Since the gray values of yellow retinal exudates and the color of the optic disc are very similar, according to the geometric characteristics of optic disc, it is necessary to apply the Hough transform first to positioning the optic disc, and then fill it. Adopting the image segmentation algorithm, the yellow retinal exudates can be extracted on the exact location in retinal images of the gray level.
With the extensive data collection, as well as the analysis and summary of existing medical retinal image processing, extract methods from hypertension lesions in non-fluorescence fundus angiography images was proposed in this paper, which can measure the diameter of the retinal vessels. According the ratio of its diameter, early diagnosis for hypertension can be finished. Meanwhile, with the accurate extraction of retinal exudates of hypertension and the clear identification between the normal and the abnormal, the standardization of hypertension research of non-fluorescence fundus angiography images can be enhanced, thus greatly strengthening the advantage of retinal examination with convenience for patients. It is also clinically significant for the diagnosis and treatment of hypertensive patients.
本文采用了Visual C++系统,在对眼底图像的特点进行分析的基础上,做了大量的实验,对非荧光造影图像进行了图像增强、边缘检测、图像分割、去噪以及视盘定位等预处理,并实现了对非荧光造影眼底图像中的动静脉血管管径的测量,根据动静脉血管的管径比可以判断该人是否为高血压病患者,并对高血压病的视网膜病变期出现的白斑病灶进行比较准确的提取,在正常和异常之间做出明确鉴别,使对高血压病眼底的非荧光造影图像的研究标准化,从而大大加强眼底检查优势,方便了高血压患者,对高血压病的诊断及治疗具有重要的实践指导意义。
The development of image processing technology has been widely applied to the scientific research, industrial and agricultural production, military technology, government departments, medical hygiene and many other fields, further promoting the development of social productivity. Today, the multidisciplinary cross-integration of digital image processing make science and technology gradually infiltrate into other areas and it is inevitable to be applied by many other subjects. Both in theory and in practice, digital image processing has the great potential in information society. Medical image processing and analysis are always the focus and difficult problems for the research of image processing and analysis. By dint of the means of graph and image technology, the displaying method and quality of medical image got a tremendous improvement, which has been greatly improved the level of diagnosis. Optic fundus has been the focus of clinical medicine because retinal lesions are closely related to the diseases of other system of the whole body. Therefore, the special signs of optic fundus can be helpful to make a correct diagnosis and prognosis estimates.
The optic fundus is "barometer" which reflect the changes of hypertensive condition. According to observing the condition of retinal arterial vessel, such as the changes of blood vessel diameter and vascular wall, the phenomenon of cross and oppression, retinal hemorrhage, exudates and its property etc., people can acquaint the whole body vascular condition Fluorescence fundus angiography is commonly used on ophthalmic clinical examination, especially for pathologic diagnosis of optic fundus and the choice of treatment. However, some side effects were reported in theapplication of this technology. Fluorescence sodium which is used as the contrast agent will result in vary effects such as nausea, vomiting, numbness or shock, cardiac arrest and other adverse reactions. Therefore, it is significant to research the means of non-Fluorescence fundus angiography. Taking account of the accuracy, objectivity, reproducibility and universality of the results of optic fundus examination, it is vital to improve the analysis of fundus images especially the quantitative analysis for retinal vessels and the extraction of retinal exudates which is closely related to hypertension. Consequently, it is essential to take effective method to segregate optic fundus images and automatically analyze and measure vascular width. Based on the changes of retinal vascular pattern, this helps ophthalmologist to make early diagnosis, monitor treatment and estimate postoperative recovery with the great clinical significance.
In this paper, base on the analysis of the characteristics of the fundus image, a lot of experiments were finished with the Visual C + + system and extract two kinds of hypertension lesions were extracted from the non-fluorescence fundus angiography images.
(1) The measurements of the blood vessel diameter in the fundus images. The optic fundus images of the hypertensive sufferer are often very unclear because of complex blood vessels, retinal hemorrhage, exudates and such lesions represent. More noise also is interferential for the follow-up identification. Therefore, the fundus images must be pretreated to reduce unrelated information for the validity, veracity of exacting eigen value. In this paper, gray linear transform in green gray image was taken to enhance the definition in the retinal RGB image, Sobel operator was applied to detect images edge and the binary image processing was performed by the settledthreshold segmentation. But there are still a lot of noise points. Subsequently, noise was reduced to process the follow-up image. After that, image quality has been improved; image features has been enhanced; Noise has been reduced, the result of image processing and analyzing has been improved and operating speed has been accelerated. As the matching pattern of retinal vascular identified system, detecting the location of the retinal optic disc is the key to identification of the retinal vessels in the optic fundus images. The border of retinal optic disc has a simple geometric shape--round features. According to the geometric characteristics of the optic disc, the Hough transform was used to detect the optic disc and it is also helpful for the next step of the measurement of vascular diameter. To reduce the amount of computation, a rectangular region was designated as the working area according to the vascular distribution in the vicinity of the optic disc. In this paper, two methods of retinal vascular diameter were applied. The first method adopted by the least square method of vascular wall for curve fitting, choose center of the optic disc which the Hough transform detect as the center to draw a circle which intersect with the vascular wall. With this way, nodal striking distance is the vessel diameter. The second method adopted the closing operation of mathematical morphology for de-noising retinal images, to fill vein first without changes in the diameter of the artery and then to measure the diameter. According to the ratio of the average artery and vein diameter, the retinal image can be judged whether it is of Hypertension retinal images. By contrast, the first method is more applicable to the detection of poor continuity of the vascular wall; the second method is adaptive for anti-jamming.
(2) The lesions extraction of retinal exudates. Since the gray values of yellow retinal exudates and the color of the optic disc are very similar, according to the geometric characteristics of optic disc, it is necessary to apply the Hough transform first to positioning the optic disc, and then fill it. Adopting the image segmentation algorithm, the yellow retinal exudates can be extracted on the exact location in retinal images of the gray level.
With the extensive data collection, as well as the analysis and summary of existing medical retinal image processing, extract methods from hypertension lesions in non-fluorescence fundus angiography images was proposed in this paper, which can measure the diameter of the retinal vessels. According the ratio of its diameter, early diagnosis for hypertension can be finished. Meanwhile, with the accurate extraction of retinal exudates of hypertension and the clear identification between the normal and the abnormal, the standardization of hypertension research of non-fluorescence fundus angiography images can be enhanced, thus greatly strengthening the advantage of retinal examination with convenience for patients. It is also clinically significant for the diagnosis and treatment of hypertensive patients.
引文
[1] 黄叔仁、张晓峰,眼底病诊断与治疗[M],北京:人民卫生出版社,2003.2:230～237
[2] 阮秋琦,数字图像处理学[M],北京:电子工业出版社,2001.10:1～2
[3] 霍宏涛、林小竹、何薇,数字图像处理[M],北京:北京理工大学出版社,2002.9:4～6
[4] 黄贤武、王加俊、李家华,数字图像处理与压缩编码技术[M],成都:电子科技大学出版社,2000.12:1
[5] 赵荣椿、赵忠明,数字图像处理导论[M],西安:西北工业大学出版社,2000.8:1～6
[6] 周长发,精通 Visual C++图像编程[M],北京:电子工业出版社,2000
[7] 杨淑莹,VC++ 图像处理程序设计[M],北京:北方交通大学出版社,2003.11,52～54
[8] 董长虹、赖志国、余啸海,Matlab 图像处理与应用[M],北京:国防工业出版社,2004.1,130～138
[9] Contrast Enhancement by Multi-scale Adaptive Histogram Equalization Yinpeng Jina,Department of Biomedical,Columbia University,New York,NY.
[10] 胡小锋、赵辉,Visual C++/MATLAB 图像处理与识别实用案例精选[M],北京:人民邮电出版社,2004.9: 85～93
[11] 王树伟、杨鸲,MATLAB 6.5 辅助图像处理[M],北京:电子工业出版社,2003.1: 184～187
[12] 段瑞玲、李庆祥、李玉和,图像边缘检测方法研究综述[J],光学技术,2005.5,31(3),415～419
[13] Kenneth R. Castleman 著,朱志刚等译,数字图像处理(新版)[M],北京:电子工业出版社,2002.2
[14] 魏弘博、吕振肃等,图像分割技术纵览[J],甘肃科学学报,2004.6 , 16(2)
[15] P. Campadelli , et al . Color image segmentation using Hopfield networks [J] . Image and Vision Computing . 1997(15) , 161～166
[16] A . Chehikian , Image segmentation by contours and regions cooperation [J] . Signal Processing , 1999(78) , 329～347
[17] Cheng H . D . , Jiang X . H . , et al . Color image segmentation: advances and prospects [J] . Pattem Recognition , 2001(34) , 2259～2281
[18] Joseph M . Reinhardt , Andrien J . Wang , et al . Cue-Based Segmentation of 4D Cardiac Image Sequences [J] . Computer Vision and Image Understanding , 2000(77) , 251～262
[19] A. Can, H. Shen , J. N. Turner , H. L. Tanenbaum , D. B. Roysam. Rapid automated tracing and feature extraction from live high-resolution retinal fundus images using direct exploratory algorithms [J]. IEEE Trans . on Info . Tech. for Biomedicine,1999,3(2): 125～138
[20] Y. A. Tolias , S. M. Panas . A fussy vessel tracking algorithm for retinal image based on fussy clustering [J]. IEEE Trans. Med. Imag . , 1998,17: 263～273
[21] S. Chaudhuri , S. Chatterjee , N. Katz , M . Nelson , M .Goldbaum . Detection of blood vessels in retinal images using two-dimensional matched filters [J]. IEEE Trans. Med. Imag. , 1989, 8(3): 263～269
[22] Chutatape, L. Zheng, S. Krishman . Retinal blood vessel detection and tracking by matched Gaussian and Kalman filters[C]. Proceeding ofIEEE Int Conf Emg and Bio Society , 1998, 20: 3144～3149
[23] X. Jiang , D. Mojon . Adaptive local thresholding by verification-based multithreshold probing with application to vessel detection in retinal images [J]. IEEE Trans. Pattern Anal. Machine Intell . 2003,25: 131～137
[24] A. Hoover , V. Kouznetsova , M. Goldmaum . Locating blood vessels in retinal images by piecewise threshold probing of a matched filter respone [J]. IEEE Trans Med Imaging,2000,19(3): 203～210
[25] T. McInerney , D. Terzopoulos . T-snakes:Topology adaptive snakes[J],Med. Imag. Anal.,2000,4(2): 73～91
[26] F. Zana , J. C. Klein . Segmentation of vessel-like patterns using mathematical morphology and curvature evaluation [J],IEEE Trans Image Processing,2001,10(7):1010～1019
[27] J. Chanussot , P. Lambert . An application of mathematical morphology to road network extraction on SAR images [J] , Mathematic Morphology and Its Application to image and signal Processing,1998: 399～406
[28] D. Thackray , A. C. Nelson . Semi-automatic segmentation of vascular network images using a rotating structuring element(rose) with mathematical morphology and dual feature thresholding [J],Patter Recogn. , 1982, 15: 431～443
[29] C. Sinthanayothin , J. F. Boyce , T. H. Williamson , et al . Automated detection of diabetic retinopathy on digital fundus images. Diabetic Med. 2002, 19(2): 105～112
[30] J. J. Staal , M. D. Abramoff , M. Niemeijer , et al . Ridge-Based Vessel Segmentation in Color Images of the Retina [J]. IEEE Transaction onMedical Imaging , 2004 , 23: 501～509
[31] Monga , N. Armande , P. Montesinos . Thin nets and crest lines: Application to satellite data and medical images [J]. Computer Vision and Image Understanding, 1997,67(3): 285～295
[32] 吴一全、朱兆达,图像处理中阈值选取方法 30 年 (1962—1992) 的进展 [J],数据采集与处理,1993,9 (3): 193～201
[33] 祝世平、周锐、申功勋,圆孔(弧)型工件的三种图像检测算法[J],仪器仪表学报,1999,20(2) : 190～192
[34] Kwang in Kim, et al . Support Vector Machines for Texture Classification[J],IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, l24(11): 1542～1550
[35] 张磊、林福宗、张钹,基于支持向量机的相关反馈图像检索算法[J],清华大学学报(自然科学版),2003,42(1):80～83
[36] P. V. C Hough. Method and Means for Recognizing Complex Patterns, U.S. Patent 3069654 , December 18 , 1962
[37] R.O. Dada and P. E. Hart , Pattern Classification and Scene Analysis , Wiley – Inter science , New York , 1973
[38] Yin P Y, Chen L H. A new method for ellipses detection using symmetry. Electronic Imaging, 1994 , 3: 20～29
[39] Leavers V F . Survey which Hough Transform CV GIP , Image Understanding, 1993, 58(2) : 250～264
[40] Yuen H K , Illingwoth J , Kitter J , Detecting partially occluded ellipses using the Hough transform , Image Vision Comput , 1989 , 7(1) : 31～37
[41] Yoo J H , Sethi I K , An ellipse detection method from the polar and pole definition of conics , Pattern Recognition , 1993 , 26(2) : 307～315
[42] Xu L , O ja E , Kultaned P . A new curve detection method : Randomized Hough Transform (RHT) , Pattern Recognition Letters , 1990 , 11(5) : 331～338
[43] 陈燕新、戚飞虎,基于随机 Hough 变换的快速圆检测方法[J],上海交通大学学报,1998.10,32(10):17～20
[44] Xu L, O ja E . Randomized Hough transform (RHT) : basic mechanisms , algorithms and computational complexities [J] , Computer Vision Graphic Image Process : Image understanding , 1993 , 57 (2) : 131～154
[45] Kalciainen H , Hirvonen P , Xu L , et a l . Probabilistic and nonprobabilistic Hough transforms : Overview and comparisons , Image and Vision Computing , 1995 , 13(4) : 239～252
[46] Kalciainen H , Hirvonen P . An extension to the randomized Hough transform exploiting connectivity , Pattern Recognition Letters , 1997 , 18 : 77～85
[47] Chen Yan-Xin , Qi Fei-Hu . A randomized Hough transform using gradient direction information , J . Infrared Millim. Waves , 1998 , 17(5) : 375
[48] G . R . J . Cooper , D . R . Cowan , The detection of circular features in irregularly spaced data [J] , Computers&Geosciences , 2004 , 101～105
[49] 林金龙、石青云,用点 Hough 变换实现圆检测的方法[J],计算机工程,2003(11),17～18、160
[50] 孙亦南、刘伟军、王越超、孙亦蓬,一种用于圆检测的改进 Hough变换方法[J],计算机工程与应用,2003.20,35～37
[51] 吴冰、秦志远,自动确定图像二值化最佳阈值的新方法[J],测绘学院学报,2001.12,18(4),283～286
[52] Donoso LA , Magaragal LE , Sanborn G , et al . Microcomputer application in retinal research [J]. Ophthalmology . 1982, 89(1): 81
[53] 曹文轩,视网膜血管测量计[J],中华眼科杂志,1964,11(1):67
[2] 阮秋琦,数字图像处理学[M],北京:电子工业出版社,2001.10:1～2
[3] 霍宏涛、林小竹、何薇,数字图像处理[M],北京:北京理工大学出版社,2002.9:4～6
[4] 黄贤武、王加俊、李家华,数字图像处理与压缩编码技术[M],成都:电子科技大学出版社,2000.12:1
[5] 赵荣椿、赵忠明,数字图像处理导论[M],西安:西北工业大学出版社,2000.8:1～6
[6] 周长发,精通 Visual C++图像编程[M],北京:电子工业出版社,2000
[7] 杨淑莹,VC++ 图像处理程序设计[M],北京:北方交通大学出版社,2003.11,52～54
[8] 董长虹、赖志国、余啸海,Matlab 图像处理与应用[M],北京:国防工业出版社,2004.1,130～138
[9] Contrast Enhancement by Multi-scale Adaptive Histogram Equalization Yinpeng Jina,Department of Biomedical,Columbia University,New York,NY.
[10] 胡小锋、赵辉,Visual C++/MATLAB 图像处理与识别实用案例精选[M],北京:人民邮电出版社,2004.9: 85～93
[11] 王树伟、杨鸲,MATLAB 6.5 辅助图像处理[M],北京:电子工业出版社,2003.1: 184～187
[12] 段瑞玲、李庆祥、李玉和,图像边缘检测方法研究综述[J],光学技术,2005.5,31(3),415～419
[13] Kenneth R. Castleman 著,朱志刚等译,数字图像处理(新版)[M],北京:电子工业出版社,2002.2
[14] 魏弘博、吕振肃等,图像分割技术纵览[J],甘肃科学学报,2004.6 , 16(2)
[15] P. Campadelli , et al . Color image segmentation using Hopfield networks [J] . Image and Vision Computing . 1997(15) , 161～166
[16] A . Chehikian , Image segmentation by contours and regions cooperation [J] . Signal Processing , 1999(78) , 329～347
[17] Cheng H . D . , Jiang X . H . , et al . Color image segmentation: advances and prospects [J] . Pattem Recognition , 2001(34) , 2259～2281
[18] Joseph M . Reinhardt , Andrien J . Wang , et al . Cue-Based Segmentation of 4D Cardiac Image Sequences [J] . Computer Vision and Image Understanding , 2000(77) , 251～262
[19] A. Can, H. Shen , J. N. Turner , H. L. Tanenbaum , D. B. Roysam. Rapid automated tracing and feature extraction from live high-resolution retinal fundus images using direct exploratory algorithms [J]. IEEE Trans . on Info . Tech. for Biomedicine,1999,3(2): 125～138
[20] Y. A. Tolias , S. M. Panas . A fussy vessel tracking algorithm for retinal image based on fussy clustering [J]. IEEE Trans. Med. Imag . , 1998,17: 263～273
[21] S. Chaudhuri , S. Chatterjee , N. Katz , M . Nelson , M .Goldbaum . Detection of blood vessels in retinal images using two-dimensional matched filters [J]. IEEE Trans. Med. Imag. , 1989, 8(3): 263～269
[22] Chutatape, L. Zheng, S. Krishman . Retinal blood vessel detection and tracking by matched Gaussian and Kalman filters[C]. Proceeding ofIEEE Int Conf Emg and Bio Society , 1998, 20: 3144～3149
[23] X. Jiang , D. Mojon . Adaptive local thresholding by verification-based multithreshold probing with application to vessel detection in retinal images [J]. IEEE Trans. Pattern Anal. Machine Intell . 2003,25: 131～137
[24] A. Hoover , V. Kouznetsova , M. Goldmaum . Locating blood vessels in retinal images by piecewise threshold probing of a matched filter respone [J]. IEEE Trans Med Imaging,2000,19(3): 203～210
[25] T. McInerney , D. Terzopoulos . T-snakes:Topology adaptive snakes[J],Med. Imag. Anal.,2000,4(2): 73～91
[26] F. Zana , J. C. Klein . Segmentation of vessel-like patterns using mathematical morphology and curvature evaluation [J],IEEE Trans Image Processing,2001,10(7):1010～1019
[27] J. Chanussot , P. Lambert . An application of mathematical morphology to road network extraction on SAR images [J] , Mathematic Morphology and Its Application to image and signal Processing,1998: 399～406
[28] D. Thackray , A. C. Nelson . Semi-automatic segmentation of vascular network images using a rotating structuring element(rose) with mathematical morphology and dual feature thresholding [J],Patter Recogn. , 1982, 15: 431～443
[29] C. Sinthanayothin , J. F. Boyce , T. H. Williamson , et al . Automated detection of diabetic retinopathy on digital fundus images. Diabetic Med. 2002, 19(2): 105～112
[30] J. J. Staal , M. D. Abramoff , M. Niemeijer , et al . Ridge-Based Vessel Segmentation in Color Images of the Retina [J]. IEEE Transaction onMedical Imaging , 2004 , 23: 501～509
[31] Monga , N. Armande , P. Montesinos . Thin nets and crest lines: Application to satellite data and medical images [J]. Computer Vision and Image Understanding, 1997,67(3): 285～295
[32] 吴一全、朱兆达,图像处理中阈值选取方法 30 年 (1962—1992) 的进展 [J],数据采集与处理,1993,9 (3): 193～201
[33] 祝世平、周锐、申功勋,圆孔(弧)型工件的三种图像检测算法[J],仪器仪表学报,1999,20(2) : 190～192
[34] Kwang in Kim, et al . Support Vector Machines for Texture Classification[J],IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, l24(11): 1542～1550
[35] 张磊、林福宗、张钹,基于支持向量机的相关反馈图像检索算法[J],清华大学学报(自然科学版),2003,42(1):80～83
[36] P. V. C Hough. Method and Means for Recognizing Complex Patterns, U.S. Patent 3069654 , December 18 , 1962
[37] R.O. Dada and P. E. Hart , Pattern Classification and Scene Analysis , Wiley – Inter science , New York , 1973
[38] Yin P Y, Chen L H. A new method for ellipses detection using symmetry. Electronic Imaging, 1994 , 3: 20～29
[39] Leavers V F . Survey which Hough Transform CV GIP , Image Understanding, 1993, 58(2) : 250～264
[40] Yuen H K , Illingwoth J , Kitter J , Detecting partially occluded ellipses using the Hough transform , Image Vision Comput , 1989 , 7(1) : 31～37
[41] Yoo J H , Sethi I K , An ellipse detection method from the polar and pole definition of conics , Pattern Recognition , 1993 , 26(2) : 307～315
[42] Xu L , O ja E , Kultaned P . A new curve detection method : Randomized Hough Transform (RHT) , Pattern Recognition Letters , 1990 , 11(5) : 331～338
[43] 陈燕新、戚飞虎,基于随机 Hough 变换的快速圆检测方法[J],上海交通大学学报,1998.10,32(10):17～20
[44] Xu L, O ja E . Randomized Hough transform (RHT) : basic mechanisms , algorithms and computational complexities [J] , Computer Vision Graphic Image Process : Image understanding , 1993 , 57 (2) : 131～154
[45] Kalciainen H , Hirvonen P , Xu L , et a l . Probabilistic and nonprobabilistic Hough transforms : Overview and comparisons , Image and Vision Computing , 1995 , 13(4) : 239～252
[46] Kalciainen H , Hirvonen P . An extension to the randomized Hough transform exploiting connectivity , Pattern Recognition Letters , 1997 , 18 : 77～85
[47] Chen Yan-Xin , Qi Fei-Hu . A randomized Hough transform using gradient direction information , J . Infrared Millim. Waves , 1998 , 17(5) : 375
[48] G . R . J . Cooper , D . R . Cowan , The detection of circular features in irregularly spaced data [J] , Computers&Geosciences , 2004 , 101～105
[49] 林金龙、石青云,用点 Hough 变换实现圆检测的方法[J],计算机工程,2003(11),17～18、160
[50] 孙亦南、刘伟军、王越超、孙亦蓬,一种用于圆检测的改进 Hough变换方法[J],计算机工程与应用,2003.20,35～37
[51] 吴冰、秦志远,自动确定图像二值化最佳阈值的新方法[J],测绘学院学报,2001.12,18(4),283～286
[52] Donoso LA , Magaragal LE , Sanborn G , et al . Microcomputer application in retinal research [J]. Ophthalmology . 1982, 89(1): 81
[53] 曹文轩,视网膜血管测量计[J],中华眼科杂志,1964,11(1):67