Average of Synthetic Exact Filters

Spring 2009

Vigniko Messan
Vijayakumar Bhagavatula
Project description

The objectives of this project are to develop and implement a tool for pattern localization. The present case is about eye localization in images. For this purpose, ASEF (Average of Synthetic Exact Filter) will be developed and implemented. ASEF is from the class of correlation filters. ASEF will basically be obtained by averaging "exact filters". "exact filters will be obtained from training images and their associated correlation output through some linear operations in the Fourier domain. Basically, an associated output is a 2-D Gaussian centered at the target. This project is included in a bigger one: face recognition. Indeed, ASEF is expected to be used to localize eyes in order to crop images for face recognition. Further on, an encryption key will be deduced. Coding will done in Mathlab. We expect ASEF to outperform other correlation filters; namely SDF, MACE, OTF, UMACE. We expect eye localization more than 98% of the time.

Return to project list