|
Abstract:
|
One of the important aspects in pattern recognition is shape classification. Given objects or contours, one would like to locate their corresponding memberships. Traditionally, one would approach this issue by taking the measurements from objects of interest or by taking the coefficients of the shape representation, namely coefficients of Fourier descriptors. Once the measurements of contours or coefficients of Fourier descriptor of contours are obtained, statistical classification procedures can be used. In our research, wavelet-based contour classification is introduced. More specifically, the wavelet transformation is applied onto the contour, and the corresponding coefficients are used for the purpose of statistical classification. The simulation study shows that the wavelet-based method is superior to the other two methods.
|
