基于改进的SIFT算法的医学显微序列图像自动配准技术
|
摘要
医学图像配准是医学图像处理领域中的一项重要技术,对临床诊断和治疗有着重要的作用。现代医学经常需要将几幅图像结合起来进行分析,以便获取更多的医疗信息。序列图像提供了比单一图像更丰富的信息,对序列图像进行分析,更有利于对目标的监测与跟踪分析。对序列图像进行分析之前首先要解决图像间对应点的严格对齐问题,也就是所谓的图像配准。医学显微序列图像的自动配准是长期以来一直未能很好解决的一个重要问题。本文主要对医学显微序列图像的自动配准技术进行研究。
提取和匹配图像特征是实现图像自动配准的一种重要方法。SIFT(尺度不变特征变换,Scale Invariant Feature Transform)最初是作为一种关键点的特征被提出来的,这种特征对图像的尺度变化和旋转变化是不变的,而且对于光照的变化和图像变形具有较强的适应性,同时,这种特征还具有较高的辨别能力,有利于后续匹配。因此,SIFT正是图像配准领域中一个热门的前沿研究方向。
本文提出一种自动的配准方法。结合医学显微序列图像的数据特点,将SIFT特征检测算法引入到显微序列图像的配准中,并针对SIFT特征描述符的高维数和高复杂度问题对特征描述符的提取方法进行了改进,降低了特征描述符的维度,简化了特征提取算法步骤,再采用双向匹配算法,剔除部分错误匹配点,最后采用随机采样一致性算法(RANSAC)去除误匹配点对以提高匹配的准确性,估计变换模型参数。统计结果表明,改进后的SIFT算法相比于传统的配准方法在保证图像配准准确度的前提下,算法复杂度和运行时间上都有了明显的降低。
Medical image registration is an important technique in the field of medical image processing. It plays a key role in clinical diagnosis and treatment. Modern medical research usually requires integrated analysis of multiple images to get more information. Sequence images can provide more abundant information than single image. Therefore, it is more available to analyze the target's movement. An initial process in sequence image integrated analysis is that the images should be perfectly registered, which is called image registration. And automatic registration in microscopic image sequence is a classical problem, which has not been solved well so far. The automatic image registration technique on microscopic image sequence has mainly been studied in this thesis.
Image feature extraction and matching is an important method to achieve automatic registration. SIFT (Scale Invariant Feature Transform) is initially proposed as a key feature, which is not variable in the scale and rotation while images being changed. It also has a strong adaptability in illumination and image deformation. In the same time, this feature also has strong recognition ability, which makes following match easier. Therefore, SIFT algorithm has attracted many attention in the field of image registration.
In this thesis, an automatic registration algorithm is presented. According to the features of medical microscopic image sequence, we introduces the SIFT feature detection method into microscopic image registration. As the dimension of the traditional SIFT descriptor is too high and the algorithm is too complex, an improved algorithm of SIFT is presented, which can reduce the dimension of SIFT descriptor. And a bidirectional matching algorithm is adopted to eliminate repeated matching points further. In addition, Random Sampling Consensus algorithm (RANSAC) is applied for removal the wrong matching points to improve the accuracy of matching further. Compared with traditional registration algorithm, the results show that the improved SIFT algorithm has been improved both in time-saving and complexity-reducing.
提取和匹配图像特征是实现图像自动配准的一种重要方法。SIFT(尺度不变特征变换,Scale Invariant Feature Transform)最初是作为一种关键点的特征被提出来的,这种特征对图像的尺度变化和旋转变化是不变的,而且对于光照的变化和图像变形具有较强的适应性,同时,这种特征还具有较高的辨别能力,有利于后续匹配。因此,SIFT正是图像配准领域中一个热门的前沿研究方向。
本文提出一种自动的配准方法。结合医学显微序列图像的数据特点,将SIFT特征检测算法引入到显微序列图像的配准中,并针对SIFT特征描述符的高维数和高复杂度问题对特征描述符的提取方法进行了改进,降低了特征描述符的维度,简化了特征提取算法步骤,再采用双向匹配算法,剔除部分错误匹配点,最后采用随机采样一致性算法(RANSAC)去除误匹配点对以提高匹配的准确性,估计变换模型参数。统计结果表明,改进后的SIFT算法相比于传统的配准方法在保证图像配准准确度的前提下,算法复杂度和运行时间上都有了明显的降低。
Medical image registration is an important technique in the field of medical image processing. It plays a key role in clinical diagnosis and treatment. Modern medical research usually requires integrated analysis of multiple images to get more information. Sequence images can provide more abundant information than single image. Therefore, it is more available to analyze the target's movement. An initial process in sequence image integrated analysis is that the images should be perfectly registered, which is called image registration. And automatic registration in microscopic image sequence is a classical problem, which has not been solved well so far. The automatic image registration technique on microscopic image sequence has mainly been studied in this thesis.
Image feature extraction and matching is an important method to achieve automatic registration. SIFT (Scale Invariant Feature Transform) is initially proposed as a key feature, which is not variable in the scale and rotation while images being changed. It also has a strong adaptability in illumination and image deformation. In the same time, this feature also has strong recognition ability, which makes following match easier. Therefore, SIFT algorithm has attracted many attention in the field of image registration.
In this thesis, an automatic registration algorithm is presented. According to the features of medical microscopic image sequence, we introduces the SIFT feature detection method into microscopic image registration. As the dimension of the traditional SIFT descriptor is too high and the algorithm is too complex, an improved algorithm of SIFT is presented, which can reduce the dimension of SIFT descriptor. And a bidirectional matching algorithm is adopted to eliminate repeated matching points further. In addition, Random Sampling Consensus algorithm (RANSAC) is applied for removal the wrong matching points to improve the accuracy of matching further. Compared with traditional registration algorithm, the results show that the improved SIFT algorithm has been improved both in time-saving and complexity-reducing.
引文
[1]陈志雄.基于图像配准的SIFT算法研究与实现[学位论文].武汉：武汉理工大学.2008：5-6页
[2]罗述谦,周国宏.医学图像处理与分析.北京：科学出版社,2003
[3]张煌,刘哲星等.医学图像信息融合技术的发展.国外医学：生物医学工程分册,2000,23(4)：202-205页
[4]吕海霞.自动图像配准技术研究[学位论文].西北工业大学,2007
[5]P. E. Anuta. Spatial registration of multispecial and multitemporal digital imagery using fast Fourier transformation techniques. IEEE Trans. on Geosci. Electron,1970, Oct, Vol. GE,8:355-368P
[6]D. I. Barnea and H. F. Silverman. A class of algorithms for fast digital registration. IEEE Trans. On Computers,1972, C-2:1179-1186P
[7]A. Arcese et al. Image detect through bipolar correlation. IEEE Trans. Inform Theory.1970, Sept.53-541P
[8]A. Rosenfield and A. C, Kak. Digital picture processing. Academic Press, Orlando, FL,1982, Vol.1 and 2:137-146P
[9]G. C. Stockman, S Kopstein, and S Benett, Matching images to models for registration and object detection via clustering. IEEE Transaction on Pattern Analysis and Machine Intelligence,1982,4:229-241P
[10]Mikio Takagi, Takamichi Hiyama, Mitsuo Sone and Morio Onoe, Automatic correcting method of geometric distortion in meteorological satellite NOAA image.电子情报通信学会论文志,1988,Vol.J71-D,No.5,883-893P
[11]Keller Y., Averbuch A. Implicit similarity a new approach to mufti-sensor image registration. Computer Vision and Pattern Recognition,2003 Proceedings,2003 IEEE Computer Society Conference.2003, June, Volume 2:18-20P
[12]R. Alexis, P. Xavier, M. Gregoire, et al. Rigid registration of 3-D ultrasound with MR images:a new approach combining intensity and gradient information, IEEE Transaction on Medical imaging,2001,20(10): 1038-1049P
[13]R.H. Huesman, GJ. Klein, J. A. Kimdon, et al. Deformable registration of multimodal data including rigid structures. IEEE Transaction on Nuclear Science,2003,50(3):389-392P
[14]P. Mark, S. Renata, M. Jacek, et al. An approach to multimodal biomedical image registration utilizing particle swarm optimization. IEEE transaction on Evolutionary Computation,2004,8(3):289-301P
[15]H. Chen, D. Yao. A MRI data processing method using projection technique. Proceeding of IEEE international conference on Robotics, Intelligent System and signal processing,2003:1235-1239P
[16]H. J. Johnson, G.E. Christensen. Consistent landmark and intensity based image registration. IEEE transaction on medical imaging,2002,21(5): 450-461P
[17]C.R. Appledorn. A new approach to the interpolation of sampled data. IEEE Transactions on Medical Imaging,1996,15:369-376P
[18]K. Rohr, H.S. Stiehl, R. Sprengel, et al. Nonrigid registration of medical images based on anatomical point landmarks and approximating thin-plate spines. Proceedings of the Aacheren Worksho,1996:41-60P
[19]D. Ruprecht, H. Muller. Free form deformation with scattered data interpolation methods. Computing Supplementum,1993,8:267-281P
[20]J. Kybic, P. Thevenaz, M. Unser. Multiresolution spine warping for EPI registration. Proceedings of the SPIE:Mathematical Imaging-Wavelet Applications in Signal and Image Processing,1999:571-579P
[21]L.G. Brown. A survey of image registration techniques. ACM Computer Surveys,1992,24(4):325-376P
[22]陈显,庄天戈.医学影像中的图像配准和融合技术研究.中国医疗器械杂志,1999,3(23)：212-217页
[23]李弼强,彭天强,彭波等.智能图像处理技术.北京：电子工业出版设.2004,239-268页
[24]周国库.基于小波提取边缘特征点的遥感图像配准技术[学位论文].西安：西北工业大学.2002:15-16页
[25]袁海军.基于特征的遥感图像自动配准研究[学位论文].成都：电子科技大学.2007：9页
[26]J.B. Antoine Maintz, P.A. van den Elsen, Max A. Viegever. Comparison of feature-based matching of CT and MR brain images. Computer Vision, Virtual Reality and Robotics in Medicine,1995:219-228P
[27]M. Fuchs, H.A. Wischman, A. Neumann, et al. Accuracy analysis for image-guided neurosurgery using fiducial skin markers,3D CT imaging and an optical localizer system. Computer Assisted Radiology,1996,1124: 770-775P
[28]W. D. Leslie, A. Borys, D. McDonald, et al. External reference markers for the correction of head rotation in brain single-photon emission tomography. European Journal of Nuclear Medicine,1995,22:351-355P
[29]Y. Amit, A. Kong. Graphical templates for model registration. IEEE Transaction on Pattern Analysis and Machine Intelligence,1996,18: 225-236P
[30]Z. Zheng, H. Wang, E. K. Teoh. Analysis of gray level corner detection. Pattern Recognition Letters,1999,20:149-162P
[31]W. Li, H. Leung. A maximum likelihood approach for image registration using control point and intensity. IEEE transaction on Image Processing, 2004,13(8):1115-1127P
[32]R, Romero, R. Rodriguez-Asomoza, J. Alarcon-Aquino, et al. Multi-modal medical image registration based on non-rigid transformations and feature point extraction by using wavelets. IEEE Instrumentation and Measurement Technology Conference Proceedings,2004:763-766P
[33]C.A Pelizzari, GT.Y. Chen, D.R. Spelbring, et al. Accurate three-dimensional registration of CT, PET, and MR images of the brain. Journal of Computer Assisted Tomography,1989,13(1):20-26P
[34]G Borgefors. Hierarchical chamfer matching:a parametric edge matching algorithm. IEEE Transaction on Pattern Analysis and Machine Intelligence, 1988,1:849-865P
[35]M.A. Audette, F.P. Ferrie, T.M. Peters. An algorithmic overview of surface registration techniques for medical imaging. Medical image Analysis,2000, 4(3):201-217P
[36]Youcef Bentoutou, Nasueddine Taleb et. An automatic image registration for applications in remote sensing. IEEE Transactions on Geosciences and Remote Sensing,2005.43:2127-2137P
[37]罗述谦.医学图像配准技术.国外医学：生物医学工程分册,1999,22(1):1-8P
[38]David G Lowe. Distinctive image features from scale-invariant keypoints, International Journal of Computer Vision,2004,60(2):91-110P
[39]Mikolajczyk K, Schmid C. A performance evaluation of local descriptors. IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27 (10):1615-1630P
[40]Yan Ke, Rahul Sukthankar. PCA-SIFT:a more distinctive representation for local image descriptors. Computer Vision and Pattern Recognition, CVPR 2004, Proceedings of the 2004 IEEE Computer Society Conference.2004, July, Vol.2:506-513P
[41]Besl P, Jain R. Segmentation and classification of range images. IEEE Transactions on Pattern Analysis and Machine Intelligence,1987,9: 608-620P
[42]I. Stamos, P E. Allen.3-D model construction using range and image data. Computer Vision and Pattern,2000 Proceedings, IEEE Conference.2000, June, Vol.1:531-536P
[43]BesI P J, McKay N D. A method for registration of 3-D shapes. IEEE Transactions on Pattern Analysis and Machine Intelligence,1992,14(2): 239-256P
[44]Y Yu, A. Ferencz, J. Malik. Extracting objects from range and radiance images. IEEE Trans, Visualization and Computer Graphics,2001,7: 352-358P
[45]Koenderink. The structure of images. Biological Cyemetics,1984,363-396P
[46]Lindeberg. Scale-space for discrete signals. IEEE Trans. PAMI,1980, 20:7-18P
[47]Jeffrey S. Beis, David G Lowe. Shape indexing using approximate Nearest-Neighbor search in high-dimensional spaces. Proceedings of the IEEE 1997 Computer Society Conference on Computer Vision and Pattern Recognition,1997:1000-1006P
[48]骞森,朱剑英..基于改进的SIFT特征的图像双向匹配算法.机械科学与技术,2007,26(9)：1179-1182P
[49]Fischler M A, Bolles R C. Random sample consensus a paradigm for model fitting with applications to image automated cartography. Communication of the ACM,1981,24(6):381-395P
[2]罗述谦,周国宏.医学图像处理与分析.北京：科学出版社,2003
[3]张煌,刘哲星等.医学图像信息融合技术的发展.国外医学：生物医学工程分册,2000,23(4)：202-205页
[4]吕海霞.自动图像配准技术研究[学位论文].西北工业大学,2007
[5]P. E. Anuta. Spatial registration of multispecial and multitemporal digital imagery using fast Fourier transformation techniques. IEEE Trans. on Geosci. Electron,1970, Oct, Vol. GE,8:355-368P
[6]D. I. Barnea and H. F. Silverman. A class of algorithms for fast digital registration. IEEE Trans. On Computers,1972, C-2:1179-1186P
[7]A. Arcese et al. Image detect through bipolar correlation. IEEE Trans. Inform Theory.1970, Sept.53-541P
[8]A. Rosenfield and A. C, Kak. Digital picture processing. Academic Press, Orlando, FL,1982, Vol.1 and 2:137-146P
[9]G. C. Stockman, S Kopstein, and S Benett, Matching images to models for registration and object detection via clustering. IEEE Transaction on Pattern Analysis and Machine Intelligence,1982,4:229-241P
[10]Mikio Takagi, Takamichi Hiyama, Mitsuo Sone and Morio Onoe, Automatic correcting method of geometric distortion in meteorological satellite NOAA image.电子情报通信学会论文志,1988,Vol.J71-D,No.5,883-893P
[11]Keller Y., Averbuch A. Implicit similarity a new approach to mufti-sensor image registration. Computer Vision and Pattern Recognition,2003 Proceedings,2003 IEEE Computer Society Conference.2003, June, Volume 2:18-20P
[12]R. Alexis, P. Xavier, M. Gregoire, et al. Rigid registration of 3-D ultrasound with MR images:a new approach combining intensity and gradient information, IEEE Transaction on Medical imaging,2001,20(10): 1038-1049P
[13]R.H. Huesman, GJ. Klein, J. A. Kimdon, et al. Deformable registration of multimodal data including rigid structures. IEEE Transaction on Nuclear Science,2003,50(3):389-392P
[14]P. Mark, S. Renata, M. Jacek, et al. An approach to multimodal biomedical image registration utilizing particle swarm optimization. IEEE transaction on Evolutionary Computation,2004,8(3):289-301P
[15]H. Chen, D. Yao. A MRI data processing method using projection technique. Proceeding of IEEE international conference on Robotics, Intelligent System and signal processing,2003:1235-1239P
[16]H. J. Johnson, G.E. Christensen. Consistent landmark and intensity based image registration. IEEE transaction on medical imaging,2002,21(5): 450-461P
[17]C.R. Appledorn. A new approach to the interpolation of sampled data. IEEE Transactions on Medical Imaging,1996,15:369-376P
[18]K. Rohr, H.S. Stiehl, R. Sprengel, et al. Nonrigid registration of medical images based on anatomical point landmarks and approximating thin-plate spines. Proceedings of the Aacheren Worksho,1996:41-60P
[19]D. Ruprecht, H. Muller. Free form deformation with scattered data interpolation methods. Computing Supplementum,1993,8:267-281P
[20]J. Kybic, P. Thevenaz, M. Unser. Multiresolution spine warping for EPI registration. Proceedings of the SPIE:Mathematical Imaging-Wavelet Applications in Signal and Image Processing,1999:571-579P
[21]L.G. Brown. A survey of image registration techniques. ACM Computer Surveys,1992,24(4):325-376P
[22]陈显,庄天戈.医学影像中的图像配准和融合技术研究.中国医疗器械杂志,1999,3(23)：212-217页
[23]李弼强,彭天强,彭波等.智能图像处理技术.北京：电子工业出版设.2004,239-268页
[24]周国库.基于小波提取边缘特征点的遥感图像配准技术[学位论文].西安：西北工业大学.2002:15-16页
[25]袁海军.基于特征的遥感图像自动配准研究[学位论文].成都：电子科技大学.2007：9页
[26]J.B. Antoine Maintz, P.A. van den Elsen, Max A. Viegever. Comparison of feature-based matching of CT and MR brain images. Computer Vision, Virtual Reality and Robotics in Medicine,1995:219-228P
[27]M. Fuchs, H.A. Wischman, A. Neumann, et al. Accuracy analysis for image-guided neurosurgery using fiducial skin markers,3D CT imaging and an optical localizer system. Computer Assisted Radiology,1996,1124: 770-775P
[28]W. D. Leslie, A. Borys, D. McDonald, et al. External reference markers for the correction of head rotation in brain single-photon emission tomography. European Journal of Nuclear Medicine,1995,22:351-355P
[29]Y. Amit, A. Kong. Graphical templates for model registration. IEEE Transaction on Pattern Analysis and Machine Intelligence,1996,18: 225-236P
[30]Z. Zheng, H. Wang, E. K. Teoh. Analysis of gray level corner detection. Pattern Recognition Letters,1999,20:149-162P
[31]W. Li, H. Leung. A maximum likelihood approach for image registration using control point and intensity. IEEE transaction on Image Processing, 2004,13(8):1115-1127P
[32]R, Romero, R. Rodriguez-Asomoza, J. Alarcon-Aquino, et al. Multi-modal medical image registration based on non-rigid transformations and feature point extraction by using wavelets. IEEE Instrumentation and Measurement Technology Conference Proceedings,2004:763-766P
[33]C.A Pelizzari, GT.Y. Chen, D.R. Spelbring, et al. Accurate three-dimensional registration of CT, PET, and MR images of the brain. Journal of Computer Assisted Tomography,1989,13(1):20-26P
[34]G Borgefors. Hierarchical chamfer matching:a parametric edge matching algorithm. IEEE Transaction on Pattern Analysis and Machine Intelligence, 1988,1:849-865P
[35]M.A. Audette, F.P. Ferrie, T.M. Peters. An algorithmic overview of surface registration techniques for medical imaging. Medical image Analysis,2000, 4(3):201-217P
[36]Youcef Bentoutou, Nasueddine Taleb et. An automatic image registration for applications in remote sensing. IEEE Transactions on Geosciences and Remote Sensing,2005.43:2127-2137P
[37]罗述谦.医学图像配准技术.国外医学：生物医学工程分册,1999,22(1):1-8P
[38]David G Lowe. Distinctive image features from scale-invariant keypoints, International Journal of Computer Vision,2004,60(2):91-110P
[39]Mikolajczyk K, Schmid C. A performance evaluation of local descriptors. IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27 (10):1615-1630P
[40]Yan Ke, Rahul Sukthankar. PCA-SIFT:a more distinctive representation for local image descriptors. Computer Vision and Pattern Recognition, CVPR 2004, Proceedings of the 2004 IEEE Computer Society Conference.2004, July, Vol.2:506-513P
[41]Besl P, Jain R. Segmentation and classification of range images. IEEE Transactions on Pattern Analysis and Machine Intelligence,1987,9: 608-620P
[42]I. Stamos, P E. Allen.3-D model construction using range and image data. Computer Vision and Pattern,2000 Proceedings, IEEE Conference.2000, June, Vol.1:531-536P
[43]BesI P J, McKay N D. A method for registration of 3-D shapes. IEEE Transactions on Pattern Analysis and Machine Intelligence,1992,14(2): 239-256P
[44]Y Yu, A. Ferencz, J. Malik. Extracting objects from range and radiance images. IEEE Trans, Visualization and Computer Graphics,2001,7: 352-358P
[45]Koenderink. The structure of images. Biological Cyemetics,1984,363-396P
[46]Lindeberg. Scale-space for discrete signals. IEEE Trans. PAMI,1980, 20:7-18P
[47]Jeffrey S. Beis, David G Lowe. Shape indexing using approximate Nearest-Neighbor search in high-dimensional spaces. Proceedings of the IEEE 1997 Computer Society Conference on Computer Vision and Pattern Recognition,1997:1000-1006P
[48]骞森,朱剑英..基于改进的SIFT特征的图像双向匹配算法.机械科学与技术,2007,26(9)：1179-1182P
[49]Fischler M A, Bolles R C. Random sample consensus a paradigm for model fitting with applications to image automated cartography. Communication of the ACM,1981,24(6):381-395P