• Open Source Computer Vision Library

主动外观模型(AAM)

Wikipedia,自由的百科全书

历史

AAM的思想最早可以追溯到1987年kass等人提出的snake方法,主要用于边界检定与图像分割。该方法用一条由n个控制点组成的连续闭合曲线作为snake模型,再用一个能量函数作为匹配度的评价函数,首先将模型设定在目标对象预估位置的周围,再通过不断迭代使能量函数最小化,当内外能量达到平衡时即得到目标对象的边界与特征。 1989年yuille等人此提出使用参数化的可变形模板来代替snake模型,可变形模板概念的提出为aam的产生奠定了理论基础。 1995年cootes等人提出的asm算法是aam的直接前身,asm采用参数化的采样形状来构成对象形状模型,并利用pca方法建立描述形状的控制点的运动模型,最后利用一组参数来控制形状控制点的位置变化从而逼近当前对象的形状,该方法只单纯利用对象的形状,因此准确率不高. 1998年,cootes等人在asm算法的基础上首先提出aam,与asm的不同之处是他不仅利用了对象的形状信息而且利用了对象的纹理信息。

AAM资源大全

websites:

1) Tim Cootes in University of Manchester, UK

http://www.isbe.man.ac.uk/~bim/

2) Mikkel B. Stegmann in Danmark Technical University, Danmark

http://www2.imm.dtu.dk/~aam/

3) Surrey Ravl -- A Fantastic Open Source for Recognition and Vision Library

http://www.ee.surrey.ac.uk/Research/VSSP/RavlDoc/share/doc/RAVL/Auto/Basic/Tree/Ravl.API.Images.AAM.html

4) Iain Matthews and Simon Baker in Carnegie Mellon University

http://www.ri.cmu.edu/project_lists/index.html

http://www.ri.cmu.edu/projects/project_448.html

5) AAMToolbox -- matlab toolbox for AAM, Dr. Andrew Hanna, University of East Anglia

http://www2.cmp.uea.ac.uk/~aih/

6) Relate -- Andrew Blake in Cambridge (Active Contour Model)

http://www.robots.ox.ac.uk/~contours/

7) Candide -- CANDIDE - a parameterized face

http://www.lysator.liu.se/~eru/research/

http://www.icg.isy.liu.se/candide/

http://www.icg.isy.liu.se/candide/javacandide.html

8) http://www.visionopen.com/cv/aam.php

9) ASMLibrary + AAMLibrary (based on opencv)

http://code.google.com/p/asmlibrary/

http://code.google.com/p/aam-library/

10) Active Shape Models with Stasm

http://www.milbo.users.sonic.net/stasm/

Publications:

1) ACM M. Kass, A. Witkin, D. Terzopoulos. "Snakes: Active Contour Models". 1st International Conference On Computer Vision, pp. 259-268, IEEE Computer Society Press, 1987.

2) ASM T. Cootes, C. Taylor, D. Cooper, J. Graham. "Active Shape Models - Their Training and Application". Computer Vision and Image Understanding, Vol. 61, No. 1, pp. 38-59, January 1995.

3) AAM T. Cootes, G. Edwards, C. Taylor. "Active Appearance Models". 5th European Conference on Computer Vision, Vol. 2, pp. 484-498, Springer, Freiburg, Germany, 1998.

4) Lucas-Kanade Fitting http://www.ri.cmu.edu/projects/project_515.html

code:

Open Source -- AAM-API, a C++ based open source developed by Dr. Mikkel B. Stegmann, Technical University of Denmark (DTU), is partly based on the open source LAPACK and Microsoft VisionSDK (now already abandoned and unavailable under Linux as well).

CMU -- CMU Robotics Institute is carrying out the project AAM Fitting Algorithms, which achieved real-time outcomes by developing Inverse Compositional Image Alignment algorithm described in "Active Appearance Models Revisited". However, CMU only affords a Matlab source code online.

Vision Open -- Vision Open has already finished rewriting AAM Building source code in C++, which is compatible with the most widely used open source libraries OpenCV and BOOST.

http://code.google.com/p/aam-library/ 写的非常规范的代码,实现了ECCV98的算法,具有很好的学习价值,在这儿也感谢这位前辈能将代码开源!

Views
Personal tools