الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 125 |  |  |
| | | from | 125 | | |
Abstract
This thesis treats image and signal analysis. In this thesis we study the properties of wavelet transforms and fractal features and its uses in signal and image analysis. We use wavelet and fractal dimension to construct an expert system to diagnose the heart mitral valve. We transform the signal by using the wavelet transform at some levels of decomposition, and then we compute the fractal dimension for the terminal nodes to create a feature vector. By using the back-propagation neural network, we classify the DHS signals into two classes, patient and fit persons.
Finally, thesis describes a method which will be used to segment the retinal blood vessel images. The method includes wavelet analysis, Gaussian mixture model, expectation maximization algorithms and supervised classifier probabilities. The method produces segmentations by classifying each image pixel as vessel or non-vessel, based on the pixel’s feature vector and uses a Bayesian classifier with class conditional probability density function described as Gaussian mixtures.