Kernel PCA and Nonlinear ASM
As a nonlinear Principal Component Analysis (PCA) method, Kernel PCA (KPCA) can effectively extract nonlinear feature. For the object image which includes more nonlinear features, traditional Active Shape Model (ASM) couldn’t obtain a good result of localization. Concerning this, an extending research on nonlinear-ASM is brought here, and an algorithm of object localization based on nonlinear-ASM is proposed. In the research of nonlinear-ASM, the problem of high dimensionality caused by nonlinear mapping has been solved effectively by the kernel theory. Besides, KPCA can not reconstruct the pre-image of the input space, thus prior model is hardly constructed by the method of the nonlinear-ASM. For solving this problem, the theory of multi-dimensional scaling is researched in the paper. The validity of the proposed method is demonstrated by the results of experiments.
KeywordsKernel Principal Component Analysis Multi-dimensional Scaling Active Shape Model Nonlinear Object Localization
Unable to display preview. Download preview PDF.
- 2.Rosipal, R., Girolami, M., Trejo, L.J.: Kernel PCA for Feature Extraction and De-noising in Non-linear Regression.Technical Report No.4, Department of Computing and Information Systems, University of Paisley (2000)Google Scholar
- 3.Cox, T.F., Cox, M.A.A.: Multidimensional Scaling, 2nd edn. Monograghs on Statistics and Applied Probability, vol. 88. Chapman & Hall/CRC (2001)Google Scholar
- 7.Williams, C.K.I.: On a Connection between Kernel PCA and Metric Multidimensional Scaling. In: Leen, T.K., Dietterich, T.G., Tresp, V. (eds.) Advances in Neural Information Processing Systems, vol. 13, pp. 675–681. MIT Press, Cambridge (2001)Google Scholar