鱼眼图像校正算法研究
|
摘要
鱼眼镜头具有短焦距(f =6~16mm)、大视场的优点(视场角约为至),在虚拟实景、视频监控、智能交通、机器人导航等领域得到广泛的应用。但鱼眼镜头摄像机拍摄的图像具有非常严重的变形,如果要利用这些具有严重变形图像的投影信息,需要将这些变形的图像校正为符合人们视觉习惯的透视投影图像。因此,对鱼眼镜头捕获的图像进行校正是一个很有意义的研究性课题。180 02700
在鱼眼图像校正之前,需要把鱼眼图像的有效区域提取出来,它在整个校正过程中至关重要。针对几种常用的有效区域提取方法的不足,本文提出一种改进算法,实验结果表明该方法在保证精度的前提下兼顾了效率。鱼眼图像校正分为2D和3D校正。2D鱼眼图像校正直接把鱼眼图像中的像素点映射到校正图像中;3D鱼眼图像校正是鱼眼图像中的像素点投影到空间中再根据校正模型将其映射到校正图像中。
本文的主要工作如下:
(1)对有关鱼眼相关文献进行了分析和总结,阐述课题研究背景及现状。讨论了图像校正相关理论知识,如:各种坐标系、针孔成像、透视投影、各种畸变和图像的预处理,针对鱼眼成像特点给出其相应的成像模型。
(2)鱼眼图像圆形有效区域提取在其整个校正中至关重要,本文给出几种常用的鱼眼图像的圆形有效区域的提取算法并给出实验结果,从适应性、复杂度、计算量的角度指出其中缺点。
(3)针对面积统计法、逐行逐列扫描法、区域生长法的缺点,本文提出了一种改进的鱼眼图像圆形有效区域提取算法并进行相应的验证,此外对这几种算法进行比较。
(4)根据鱼眼图像成像特点,分别从2D和3D空间的角度校正鱼眼图像——经度坐标校正、等距投影校正、球面约束模型校正。
As fish-eye lens have the advantages of short focal length (f=6~16mm) and large field (about even ), Thus it have been very widely used in many areas, such as video conferencing, video surveillance, intelligent transportation, panoramic view etc. But the fish-eye images that are taken with the camera has a very serious distortion, if we want to use the information of projection image with serious distortion, these deformed images need to be corrected to comply with the visual habits of the perspective projection image. Therefore, Correcting the images that is captured by fish-eye lens is worthy of the subject of research. 180 02700
The effective area need to be extracted before the fish-eye image correction. It is critical in the entire correction process. In this paper, an improved algorithm is given for the deficiencies of several commonly used extraction methods. The experimental results show that the method improves efficiency under ensuring the premise of accuracy. The fish-eye image correction is divided into 2D and 3D correction. 2D fish-eye image correction directly map the pixel of the fish-eye image to the corrected image, 3D fish-eye image correction is that the pixel in the image is projected into three dimension space and then it is mapped to the corrected image by corrected model.
The main works of the paper are as follows:
(1) The relevant literatures about the fish-eye is analyzed and summarized, that describes the development of fish-eye image technique and the significance of the fish-eye image correction. Discussion of the relevant theoretical knowledge of image correction, such as: all kinds of coordinates, the pinhole image, perspective projection, all kinds of distortion and image preprocessing. The image model is given by the feature of fish-eye image.
(2) Fish-eye image extraction circular effective area is very important throughout the whole correction. This paper gives some usually used the extraction algorithms of the effective circular area for fish-eye image and experimental results are given. The paper points out the shortcomings from the adaptability, complexity, computation.
(3) As the shortcoming of area statistics, sweeping line by line and region growing, this paper gives the improvement algorithm of the effective circular area for fish-eye image and the corresponding verification. In addition, it compares several algorithms.
(4) According to the characteristic of fish-eye image, fish-eye image are corrected respectively from the perspective of 2D and 3D space, such as: longitude coordinate correction, equidistance projection and spherical constraint model. Finally, verifying the algorithms.
在鱼眼图像校正之前,需要把鱼眼图像的有效区域提取出来,它在整个校正过程中至关重要。针对几种常用的有效区域提取方法的不足,本文提出一种改进算法,实验结果表明该方法在保证精度的前提下兼顾了效率。鱼眼图像校正分为2D和3D校正。2D鱼眼图像校正直接把鱼眼图像中的像素点映射到校正图像中;3D鱼眼图像校正是鱼眼图像中的像素点投影到空间中再根据校正模型将其映射到校正图像中。
本文的主要工作如下:
(1)对有关鱼眼相关文献进行了分析和总结,阐述课题研究背景及现状。讨论了图像校正相关理论知识,如:各种坐标系、针孔成像、透视投影、各种畸变和图像的预处理,针对鱼眼成像特点给出其相应的成像模型。
(2)鱼眼图像圆形有效区域提取在其整个校正中至关重要,本文给出几种常用的鱼眼图像的圆形有效区域的提取算法并给出实验结果,从适应性、复杂度、计算量的角度指出其中缺点。
(3)针对面积统计法、逐行逐列扫描法、区域生长法的缺点,本文提出了一种改进的鱼眼图像圆形有效区域提取算法并进行相应的验证,此外对这几种算法进行比较。
(4)根据鱼眼图像成像特点,分别从2D和3D空间的角度校正鱼眼图像——经度坐标校正、等距投影校正、球面约束模型校正。
As fish-eye lens have the advantages of short focal length (f=6~16mm) and large field (about even ), Thus it have been very widely used in many areas, such as video conferencing, video surveillance, intelligent transportation, panoramic view etc. But the fish-eye images that are taken with the camera has a very serious distortion, if we want to use the information of projection image with serious distortion, these deformed images need to be corrected to comply with the visual habits of the perspective projection image. Therefore, Correcting the images that is captured by fish-eye lens is worthy of the subject of research. 180 02700
The effective area need to be extracted before the fish-eye image correction. It is critical in the entire correction process. In this paper, an improved algorithm is given for the deficiencies of several commonly used extraction methods. The experimental results show that the method improves efficiency under ensuring the premise of accuracy. The fish-eye image correction is divided into 2D and 3D correction. 2D fish-eye image correction directly map the pixel of the fish-eye image to the corrected image, 3D fish-eye image correction is that the pixel in the image is projected into three dimension space and then it is mapped to the corrected image by corrected model.
The main works of the paper are as follows:
(1) The relevant literatures about the fish-eye is analyzed and summarized, that describes the development of fish-eye image technique and the significance of the fish-eye image correction. Discussion of the relevant theoretical knowledge of image correction, such as: all kinds of coordinates, the pinhole image, perspective projection, all kinds of distortion and image preprocessing. The image model is given by the feature of fish-eye image.
(2) Fish-eye image extraction circular effective area is very important throughout the whole correction. This paper gives some usually used the extraction algorithms of the effective circular area for fish-eye image and experimental results are given. The paper points out the shortcomings from the adaptability, complexity, computation.
(3) As the shortcoming of area statistics, sweeping line by line and region growing, this paper gives the improvement algorithm of the effective circular area for fish-eye image and the corresponding verification. In addition, it compares several algorithms.
(4) According to the characteristic of fish-eye image, fish-eye image are corrected respectively from the perspective of 2D and 3D space, such as: longitude coordinate correction, equidistance projection and spherical constraint model. Finally, verifying the algorithms.
引文
[1]吕宏静.用于星球探测机器人导航的立体视觉技术.北京理工大学硕士学位论文,2000年2月.
[2]张诚,汪嘉业.对鱼眼照片场景实现三维重建和虚拟浏览[J].计算机辅助设计与图形学报, 2004, 16(1):79-84.
[3] Smith W E, Vakil N and Maislin S A. Correction of distortion in endoscope images [J]. IEEE Trans. on Medical Imaging, 1992, 11:117-122.
[4] Huaibin Zhao. Agganval. 3D Reconstruction of an Urban Scene from Synthetic Fish-eye Images [C]. IEEE Proc.of image Analysia and Interpretation, Anstin, TX, U.S.A , 2000: 219-223.
[5] R G Yang, M S Brownt, W B Sealest, Geometrically Correct Imagery for Teleconferencing, In: ACM Multimedia’99, Orlando, Nov 1st, 1999, 179-186.
[6]王永仲.鱼眼镜头光学[M].北京:科学出版社,2008.
[7]孙丽华.基于鱼眼镜头的全视觉图像研究.哈尔滨工程大学硕士论文,2008年7月.
[8]英向华,胡占义.一种基于球面透视投影约束的鱼眼镜头校正方法[J].计算机学报, 2003, 26(12):1702-1708.
[9]王俊杰,徐小刚,胡运发,于玉.鱼眼投影在虚拟实景中的应用研究.小型微型计算机系统[J], 2004, 25(2):1-7.
[10]贾云得,吕宏静,刘万春.鱼眼变形立体图像恢复稠密深度图的方法.计算机学报[J], 2003, 23(12):1332-1336.
[11]张诚,汪嘉业.对鱼眼照片场景实现三维重建和虚拟浏览.计算机辅助几何设计与图形学报[J], 2004, 16(1):79-84.
[12] T.N.Mundhenk, M.J.Rivett, Techniques for Fish-eye Lens Calibration using a Minimal Number of Measurements, Procof the SPIE Intelligent Robotics and Computer Vision Conference, Boston, Massachusetts, 2000, vol.11, pp: 8-9.
[13] Y.Nomura, M.Sagara, Hiroshi Naruse. Simple calibration algorithm for high-distortion-lens camera [J]. IEEE Transactions on Pattern Analysis and Machine Intelligecne, 1929, 14(11): 1095-1019.
[14]邱志强,陆宏伟.于起峰.用射影不变性纠正鱼眼镜头畸变[J].应用光学,2003, 24(5): 36-38.
[15]廖士中.高培焕.一种光学镜头摄像机图像几何畸变的修正方法[J].中国图像图形学报, 2000, 5(7):593-596.
[16]周海林,王立琦.光学图像几何畸变的快速校正算法[J].中国图像图形学报,2003, 8(10): 1131-1135.
[17] S.Shah, J.K.Aggarwal. A Simple Calibration Procedure for Fish-Eye (High Distortion) Lens Camera[C]. In: Proceedings of the IEEE International Conference on Robotics and Automation, San Diego, CA, USA, 1994, 3: 3422-3427.
[18] A.Beau, S.Licardie, Modeling fish-eye lenses[C], In: Proceedings of IEEE International Conference on Intelligent Robots and Systems Yokohama, 1993, 6: 1822-1828.
[19] S.S.Beauchemin, R.Bajcsy. Modelling and Removing Radial and Tangential Distortion in Spherical Lenses, Multi-Image Analysis, LNCS 2032, 2001: 1-21.
[20] G.Glaeser, E.Groller. Fast generation of curved perspectives for ultra-wide-angle lenses in VR applications [J]. The Visual Computer, 1999, 15:365-376.
[21] J.Kannala, S.S.Brandt. A generic camera model and calibration method for conventional, wide-angle and fish-eye lenses [J]. IEEE Trans. Pattern Anal Machine Intell PAMI, 2006, 8:1335-1340.
[22] H.Bakstein. Calibration of a Fish-Eye Lens with Field of View Larger than 180 [C]. Proceedings of the CVWW2002, February 2002, pp:276-285.
[23] X.H Ying, ZY Hu. Can We Consider Central Catadioptric Cameras and Fish-eye Cameras within a Unified Imaging Model [C]. In Proc. European Conference on Computer Vision, 2004, 3021:442-445.
[24] J.Kanala, S.Brandt. A Generic Camera Calibration Method for Fish-eye Lenses [C]. In: Proceedings of the l7th Conference on Pattern Recognition, 2003, 1:10-13.
[25] D.B.Gennery. Generalized Camera Calibration Including Fish-Eye Lenses International Journal of Computer Vison, 2006, 68(3):239-226.
[26] Shah S, Aggarwal J K. Intrinsic parameter calibration procedure for a (high distortion) fish-eye lens camera with distortion model and accuracy estimation [J]. Pattern Recognition, 1996, 29(11):1775-1788.
[27]黄友度,苏化明.一种鱼眼图像到透视投影图像的变换模型[J].系统仿真学报, 2005, 17(1):29-32.
[28]张广军.机器视觉[M].北京科学出版社, 2005年6月.
[29] http://www.nlpr.ia.ac.cn/English/rv/download.htm
[30]马颂德,张正友.计算理论与算法基础[M].北京:科学出版社, 1998.
[31]汪嘉业,杨兴强,张彩明.基于鱼眼镜头拍摄的图像生成漫游模型[J].系统仿真学报, 2001, 11(13):66-68.
[32] Maybank S. Theory of reconstruction from image motion [M]. New York: Springer Series in Information Sciences, Springer-Verlag, 1993.
[33]唐俊.鱼眼图像轮廓提取算法研究[J ].微机发展, 2004, 14(10):10-12.
[34]唐俊,赵为民,谷峰.基于鱼眼图像的全景漫游模型[J].微机发展, 2003, 13(2):69-72.
[35]王大宇,崔汉国,陈军.鱼眼图像轮廓提取及校正研究[J].计算机工与设计, 2007, 28(6): 2878-2879.
[36]崔汉国,陈军,王大宇.基于鱼眼图像的虚拟漫游研究[J].系统仿真学报, 2007, 9(5) :1504-1506.
[37]杨丹,游磊,张小洪,李博.基于区域生长的鱼眼图像轮廓提取算法[J].计算机工程,2010, 36(8):217-218.
[38]陈兵旗,孙明. Visual C++实用图像处理[M].北京:清华大学出版社, 2004.
[39]周辉,罗飞,李慧娟,冯炳枢.基于柱面模型的鱼眼影像校正方法的研究[J].计算机应用, 2008,10(28): 2264-2266.
[40]邱志强,陆宏伟,于起峰.用摄影不变性纠正鱼眼镜头畸变[J].应用光学, 2003, 24(5): 36-38.
[41] Ahemd M T, Hemayed E E, Farag A A. A camera calibration with lens distortion using weakly-localized image feature [C]. In: Proceedings of the SPI, 1999, 3650:119-127.
[42] R.Y.Tsai. A versatile camera calibration technique for high-accuracy 3D machine vision metrology using off-the-shelf TV cameras and lenses [J]. IEEE Journal of Robotics and Automation, 1987, 3(4):323-344.
[43]贾云得,吕宏静,徐岸.一种鱼眼镜头成像立体视觉系统的标定方法[J].计算机学报, 2000, 23(11): 1215-1219.
[44] Y.Xiong, K.Turkowski. Creating Image-Based VR Using a Self-Calibrating Fish-eye Lens [C]. In:Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition.San Juan, Puerto Rico, USA, 1997:237-243.
[45] Swaminathan R, Nayar S K. Non-metric calibration of wide-angle lenses and poly cameras [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(10): 1172-1178.
[46] Kang S B. Radial distortion snakes [J]. In: IAPR Workshop on Machine Vision Applications, Tokyo, 2000:603-606.
[47] Devernay F, Faugeras O. Straight lines have to be straight: automatic calibration and removal of distortion from scenes of structured environments [J]. Machine Vision and Applications, 2001, 14-24.
[48] Weng J, Cohen P and Herniou M. Camera calibration with distortion models and accuracy evaluation [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(10): 965-980.
[49] Nomura Y, Sagara M, Naruse H. A Simple calibration algorithm for high-distortion-lens camera [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(11): 1095-1100.
[50] Zhang Z. On the epipolar geometry between two images with lens distortion [C]. In: Proceedings of the International Conference on Pattern Recognition, 1996, 407-411.
[51] Stein G P. Lens distortion calibration using point correspondences [C]. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 1997, 143-148.
[52] Fitzgibbon A W. Simultaneous linear estimation of multiple view geometry and lens distortion [C]. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2001:125:132.
[2]张诚,汪嘉业.对鱼眼照片场景实现三维重建和虚拟浏览[J].计算机辅助设计与图形学报, 2004, 16(1):79-84.
[3] Smith W E, Vakil N and Maislin S A. Correction of distortion in endoscope images [J]. IEEE Trans. on Medical Imaging, 1992, 11:117-122.
[4] Huaibin Zhao. Agganval. 3D Reconstruction of an Urban Scene from Synthetic Fish-eye Images [C]. IEEE Proc.of image Analysia and Interpretation, Anstin, TX, U.S.A , 2000: 219-223.
[5] R G Yang, M S Brownt, W B Sealest, Geometrically Correct Imagery for Teleconferencing, In: ACM Multimedia’99, Orlando, Nov 1st, 1999, 179-186.
[6]王永仲.鱼眼镜头光学[M].北京:科学出版社,2008.
[7]孙丽华.基于鱼眼镜头的全视觉图像研究.哈尔滨工程大学硕士论文,2008年7月.
[8]英向华,胡占义.一种基于球面透视投影约束的鱼眼镜头校正方法[J].计算机学报, 2003, 26(12):1702-1708.
[9]王俊杰,徐小刚,胡运发,于玉.鱼眼投影在虚拟实景中的应用研究.小型微型计算机系统[J], 2004, 25(2):1-7.
[10]贾云得,吕宏静,刘万春.鱼眼变形立体图像恢复稠密深度图的方法.计算机学报[J], 2003, 23(12):1332-1336.
[11]张诚,汪嘉业.对鱼眼照片场景实现三维重建和虚拟浏览.计算机辅助几何设计与图形学报[J], 2004, 16(1):79-84.
[12] T.N.Mundhenk, M.J.Rivett, Techniques for Fish-eye Lens Calibration using a Minimal Number of Measurements, Procof the SPIE Intelligent Robotics and Computer Vision Conference, Boston, Massachusetts, 2000, vol.11, pp: 8-9.
[13] Y.Nomura, M.Sagara, Hiroshi Naruse. Simple calibration algorithm for high-distortion-lens camera [J]. IEEE Transactions on Pattern Analysis and Machine Intelligecne, 1929, 14(11): 1095-1019.
[14]邱志强,陆宏伟.于起峰.用射影不变性纠正鱼眼镜头畸变[J].应用光学,2003, 24(5): 36-38.
[15]廖士中.高培焕.一种光学镜头摄像机图像几何畸变的修正方法[J].中国图像图形学报, 2000, 5(7):593-596.
[16]周海林,王立琦.光学图像几何畸变的快速校正算法[J].中国图像图形学报,2003, 8(10): 1131-1135.
[17] S.Shah, J.K.Aggarwal. A Simple Calibration Procedure for Fish-Eye (High Distortion) Lens Camera[C]. In: Proceedings of the IEEE International Conference on Robotics and Automation, San Diego, CA, USA, 1994, 3: 3422-3427.
[18] A.Beau, S.Licardie, Modeling fish-eye lenses[C], In: Proceedings of IEEE International Conference on Intelligent Robots and Systems Yokohama, 1993, 6: 1822-1828.
[19] S.S.Beauchemin, R.Bajcsy. Modelling and Removing Radial and Tangential Distortion in Spherical Lenses, Multi-Image Analysis, LNCS 2032, 2001: 1-21.
[20] G.Glaeser, E.Groller. Fast generation of curved perspectives for ultra-wide-angle lenses in VR applications [J]. The Visual Computer, 1999, 15:365-376.
[21] J.Kannala, S.S.Brandt. A generic camera model and calibration method for conventional, wide-angle and fish-eye lenses [J]. IEEE Trans. Pattern Anal Machine Intell PAMI, 2006, 8:1335-1340.
[22] H.Bakstein. Calibration of a Fish-Eye Lens with Field of View Larger than 180 [C]. Proceedings of the CVWW2002, February 2002, pp:276-285.
[23] X.H Ying, ZY Hu. Can We Consider Central Catadioptric Cameras and Fish-eye Cameras within a Unified Imaging Model [C]. In Proc. European Conference on Computer Vision, 2004, 3021:442-445.
[24] J.Kanala, S.Brandt. A Generic Camera Calibration Method for Fish-eye Lenses [C]. In: Proceedings of the l7th Conference on Pattern Recognition, 2003, 1:10-13.
[25] D.B.Gennery. Generalized Camera Calibration Including Fish-Eye Lenses International Journal of Computer Vison, 2006, 68(3):239-226.
[26] Shah S, Aggarwal J K. Intrinsic parameter calibration procedure for a (high distortion) fish-eye lens camera with distortion model and accuracy estimation [J]. Pattern Recognition, 1996, 29(11):1775-1788.
[27]黄友度,苏化明.一种鱼眼图像到透视投影图像的变换模型[J].系统仿真学报, 2005, 17(1):29-32.
[28]张广军.机器视觉[M].北京科学出版社, 2005年6月.
[29] http://www.nlpr.ia.ac.cn/English/rv/download.htm
[30]马颂德,张正友.计算理论与算法基础[M].北京:科学出版社, 1998.
[31]汪嘉业,杨兴强,张彩明.基于鱼眼镜头拍摄的图像生成漫游模型[J].系统仿真学报, 2001, 11(13):66-68.
[32] Maybank S. Theory of reconstruction from image motion [M]. New York: Springer Series in Information Sciences, Springer-Verlag, 1993.
[33]唐俊.鱼眼图像轮廓提取算法研究[J ].微机发展, 2004, 14(10):10-12.
[34]唐俊,赵为民,谷峰.基于鱼眼图像的全景漫游模型[J].微机发展, 2003, 13(2):69-72.
[35]王大宇,崔汉国,陈军.鱼眼图像轮廓提取及校正研究[J].计算机工与设计, 2007, 28(6): 2878-2879.
[36]崔汉国,陈军,王大宇.基于鱼眼图像的虚拟漫游研究[J].系统仿真学报, 2007, 9(5) :1504-1506.
[37]杨丹,游磊,张小洪,李博.基于区域生长的鱼眼图像轮廓提取算法[J].计算机工程,2010, 36(8):217-218.
[38]陈兵旗,孙明. Visual C++实用图像处理[M].北京:清华大学出版社, 2004.
[39]周辉,罗飞,李慧娟,冯炳枢.基于柱面模型的鱼眼影像校正方法的研究[J].计算机应用, 2008,10(28): 2264-2266.
[40]邱志强,陆宏伟,于起峰.用摄影不变性纠正鱼眼镜头畸变[J].应用光学, 2003, 24(5): 36-38.
[41] Ahemd M T, Hemayed E E, Farag A A. A camera calibration with lens distortion using weakly-localized image feature [C]. In: Proceedings of the SPI, 1999, 3650:119-127.
[42] R.Y.Tsai. A versatile camera calibration technique for high-accuracy 3D machine vision metrology using off-the-shelf TV cameras and lenses [J]. IEEE Journal of Robotics and Automation, 1987, 3(4):323-344.
[43]贾云得,吕宏静,徐岸.一种鱼眼镜头成像立体视觉系统的标定方法[J].计算机学报, 2000, 23(11): 1215-1219.
[44] Y.Xiong, K.Turkowski. Creating Image-Based VR Using a Self-Calibrating Fish-eye Lens [C]. In:Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition.San Juan, Puerto Rico, USA, 1997:237-243.
[45] Swaminathan R, Nayar S K. Non-metric calibration of wide-angle lenses and poly cameras [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(10): 1172-1178.
[46] Kang S B. Radial distortion snakes [J]. In: IAPR Workshop on Machine Vision Applications, Tokyo, 2000:603-606.
[47] Devernay F, Faugeras O. Straight lines have to be straight: automatic calibration and removal of distortion from scenes of structured environments [J]. Machine Vision and Applications, 2001, 14-24.
[48] Weng J, Cohen P and Herniou M. Camera calibration with distortion models and accuracy evaluation [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(10): 965-980.
[49] Nomura Y, Sagara M, Naruse H. A Simple calibration algorithm for high-distortion-lens camera [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(11): 1095-1100.
[50] Zhang Z. On the epipolar geometry between two images with lens distortion [C]. In: Proceedings of the International Conference on Pattern Recognition, 1996, 407-411.
[51] Stein G P. Lens distortion calibration using point correspondences [C]. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 1997, 143-148.
[52] Fitzgibbon A W. Simultaneous linear estimation of multiple view geometry and lens distortion [C]. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2001:125:132.