2D invariant image recognition using the reduced coefficient of elliptic Fourier and principal component analysis with neural network
Keywords:Elliptic Fourier descriptors, principal component analysis
This paper presents the methods to show how to recognize the boundary shape of 2D invariant images using the reduced coefficient of elliptic Fourier descriptors (RCEFDs) and principal component analysis (PCA) with neural networks. The boundary of each image is represented by the amount of arc (harmonic) ellipses and then to compute the coefficient of elliptic Fourier descriptors (CEFDs) to be image codes. The system can save storage size by a reduced number of coefficient codes; using of the three coefficients (an, bn, cn or an, cn, dn) instead of four coefficients (an, bn, cn, dn).The amount of harmonics which best fit to the boundary of each object is 15 harmonics. For getting a satisfactory recognition rate and for solving the problems that the objects have size variables and orientations, it is necessary to set the group of objects by rotating the original object by different degrees, normalizing the scale to standard size and translating to a suitable situation. In this paper, 100 images are used for training and 100 images are used for testing. In the Training stage, each image is rotated into 36 different angles. The principal component analysis is then applied to find the mean of each object image for storing only instead of storing all member sets into the knowledge base for marking a small-scale knowledge base. In the testing stage, the RCEFDs and PCA method are applied to find the output comparison using classical (PCA) method with the Neural Network method. This proposed method can reduce storage size of Fourier coefficients by 1/4, and then the PCA used storage size to 1/36 of conventional method. The results show that: the recognition rate of PCA with Neural Network method is 94.23%, and the recognition rate of the Neural Network method is 96.20%, respectively.
Anil, K. (1989). Fundamentals of Digital Image Processing. Prentice-Hall International Englewood Cliffs, NJ, 342-425.
Geusebroek, J. M., Burghouts, G. J. & Smeulders A. W. M. (2005) The Amsterdam library Of object images.International Journal of Computer Vision, 13-122.
Kuhl, F. P. & Giardina, C. R. (1982). Elliptic Fourier features of a closed contour. Comp Graph. Image Proc, 236-258.
Pansang, S., Attachoo, B. & Kimpan, C. (2005). Invariant Rang Image Multi-Pose Face Recognition using Gradient Face, Membership Matching Score and 3-Layer Matching Search. IEICE TRANS.INF.& SYST., Vol. E88-D, 188-179.
Phokharatkul, P., Marang, U. & Kimpan, C. (2000). Recognition of Printed Thai Characters Using Elliptic Fourier Descriptors and Genetic Neural Networks. Proceedings Asia-Pacific Symposium on Broadcasting and Communications, 406-411.
Phokharatkul, P. & Kimpan, C. (2002). Handwritten thai character recognition using Fourier descriptors and genetic neural network. Comput.Intell, 270–293.
Phokharatkul, P., Kamnuanchai, S. & Kimpan, C. (2006). Invariant Shape Object Recognition Using B-Spline, Cardinal Spline and Genetic Algorithm. WSEAS on Signal Processing, 40-46.
Phokharatkul, P., Kamnuanchai, S. & Kimpan, C. (2007). Invariant Image Recognition Using Elliptic Fourier Descriptors and Membership Matching Score. Proc, International Workshop on Advanced Image Technology, 26.
Richard, D. & Peter, E. (1973) Contour Following. Pattern Classification and Scene Analysis. Wiley-Interscience, 290–293.
Zhu, H., Liu, M. & Ji, H. (2010). Combined invariants to blur and rotation using Zernike moment descriptors. Pattern Analysis Appliations, Sringer, 13, 309-319.
How to Cite
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.