The objective of this research is to develop a classifier that can reliably and accurately discriminate among a large number of different natural-surfaces in color images using only general-purpose color and texture features. The general-purpose color and texture features are those which exhibit the least sensitivity to illumination and viewpoint variation in a broad range of applications for which color and texture are a reasonable basis for classification. The feature probability density function (PDF) distributions of natural-surface classes are most definitely not disjoint, which instantly obviates the appropriate use of Bayes' Decision Rule, which is profoundly confused by classes that are heavily overlapped in feature space. Therefore, it was also necessary to develop a new ALISA δCRC classifier that would not be confused by the many different and often highly similar natural-surface classes. A series of experiments have been designed and conducted using the CUReT image database, Caltech facial photo images, and natural images. An ALISA δ CRC classifier was trained with up to 61 classes in the CUReT image database, which presents each class in 205 different and carefully controlled viewpoint and illumination conditions. The results with images not in the training set yielded classification accuracies well above 95%. This research then extended the classifier feasibility tests and accuracy measurement using mosaic images made up from different CUReT material surface classes. Next, a skin classification experiment was conducted using Caltech facial photo images that had a single class of interest with complex backgrounds. Finally, a natural image-content classification experiment was performed using natural scenery images that had several classes of interest and irregular class boundaries.
|Commitee:||Florea, Liliana, Happel, Mark, Martin, Dianne, Youssef, Abdou|
|School:||The George Washington University|
|School Location:||United States -- District of Columbia|
|Source:||DAI-B 69/07, Dissertation Abstracts International|
|Keywords:||Adaptive learning, Classification methodology, Color and texture features, Machine vision, Statistical pattern recognition, Viewpoint and illumination invariance|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be