Facial image processing - eyes extraction and glasses matching

Abstract

Eye detection is not new to computer industry. It has been widely used in applications, for example red eye detection, eye graze detection, eye recognition, human-computer interaction system etc. Besides, it is a critical task of face detection as well. This project provides an alternative approach to develop an eye detection algorithm that is built on feature based eye detection methods. The new algorithm uses the image‟s pixel color as the only information for eye tracking. It is straightforward and does not contain any heavy processing procedures such as template matching or the implementation of crummy training data. Moreover, the project brings the suggested eye detection algorithm to life application - Eyewear Matching System. This system is to show how the algorithm can help to locate the eyes and merge an eyewear photo to the eyes position. As merging two images together may cause aliasing, merging techniques – erosion and Gaussian smoothing are included in this project.