An improved eye feature extraction algorithm based on deformable templates

In this paper, an improved method for eye extraction using deformable templates is proposed. This new method overcomes the shortcomings of traditional deformable template techniques for eye extraction, such as unexpected shrinking of the template and complexity of the updating procedure, while offering higher flexibility and accuracy. A new size term and eye corner finder are introduced to prevent over-shrinking and improve speed and accuracy of fitting. The eye features are fitted in a pre-set order to reduce the complexity of the updating procedure and increase the flexibility. We demonstrate the success of this new method by extracting eye features in real images.

• Type: Conference
• Year: 2005
• Where: ICIP
• Authors: Paul Kuo, John Hannah
• Keywords: Feature Extraction Algorithm, ICIP 2005, Image Processing