الفهرس Only 14 pages are availabe for public view
 |  | from | 72 |  |  |
| | | from | 72 | | |
Abstract
The aim of this thesis is to develop a system for the automatic recognition of hieroglyphic symbols. Only twenty hieroglyphic symbols were considered in this thesis. Two main concerns were addressed. Choosing the best features in terms of recognition performance, low computational complexity, and suitability for real time applications. The other concern was the choice of classifier.
Two types of features were examined. In the first type, features were extracted from the standard discrete Wavelet transform (DWT) of the image. Four Wavelet families were employed: re{l.\ Daubechies wavelets, biorthogonal wavelets, complex symmetric Daubechies wavelets, and the dual-tree complex Wavelet transform (DT-CWT).
The second type of features was obtained using moment-based shape descriptors.
Features were extracted using .Zernike moments (ZM), Chebyshev moments (CM), and Radial Chebyshev moments (RCM).
The candidate features were used to train backpropagation neural networks. The trained networks were tested to compare the performance of the different sets of features. The features were also used to design an adaptive neuro-fuzzy inference system (ANFIS) for classification, and the performance of the designed neuro-fuzzy classifiers was evaluated.
The results showed that the use of Chebyshev moments for feature extraction, combined with the use of neuro-fuzzy classifiers, yielded the best recognition rate with the least number of features.