Image Processing and Neural Networks Lab

Reference Texts:  Statistical Pattern Recognition by Keinosuke Fukunaga

                            Pattern Classification by Duda, Hart, and Stork

Professor: M.T. MANRY 

Currently, there is an increasing demand for graduates with skills in the areas of machine learning and statistical pattern recognition. In this course, we investigate algorithms used to construct pattern recognition systems for applications such as automatic document reading (as in zip code recognition), face recognition, fingerprint recognition, multimedia wireless, prognostics, bioinformatics, automatic target recognition and robotic vision. Program assignments will be given that are based upon a character recognition application, using image files and image reading software that are provided.

In the first part of the course we will investigate methods for calculating deformation-invariant and deformation-variant feature vectors from waveforms and imagery. Classical feature sets such as moments and Fourier descriptors will be covered, along with general transform-based approaches. 

In the second part of the course, we will study decision theory. Conventional Bayes classifiers and nearest neighbor classifiers will be covered. Classifiers designed through regression such as the piecewise linear and neural network classifiers will be investigated.  

In the last part of the course we will study decision fusion approaches for merging classifiers, including boosting algorithms. Subsetting approaches for feature selection will be covered including floating search and branch and bound. Prerequisites: knowledge of digital signal/image processing or random processes, image processing.

Prerequisites: EE5352 or knowledge of digital signal processing or digital image processing.

Click here to email any questions or comments about this web site.

Copyright © 2004 Image Processing and Neural Networks Lab
Last modified: March 06, 2007