Face Recognition Using the Discrete Cosine Transform.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH SCHOOL OF ENGINEERING
Pagination or Media Count:
The purpose of this study was to improve the feature extraction capability and thus, the recognition accuracy of the AFIT Face Recognition Machine AFRM. The discrete cosine transform DCT was analyzed in depth to determine its image compression and feature extraction capabilities. Features were extracted using a whole image and using sub-blocks of an image. The features extracted were tested for recognition accuracy using a nearest neighbor network classifier. Tests were run to determine if a single person could be distinguished from multiple individuals. Tests were also run to determine if the net could discriminate between multiple individuals. The results were compared with the discrete Fourier transform DFT and the Karhunen-Loeve transform KLT. The DCT results were superior in all cases to those obtained with the DFT and in some case were even superior to those obtained with the KLT.