An efficient iris and eye corners extraction method

Abstract : Eye features are one of the most important clues for many computer vision applications. In this paper, an efficient method to automatically extract eye features is presented. The extraction is highly based on the usage of the common knowledge about face and eye structure. With the assumption of frontal face images, firstly coarse eye regions are extracted by removing skin pixels in the upper part of the face. Then, iris circle position and radius are detected by using Hough transform in a coarse-to-fine fashion. In the final step, edges created by upper and lower eyelids are detected and polynomials are fitted to those edges so that their intersection points are labeled as eye corners. The algorithm is experimented on the Bosphorus database and the obtained results demonstrate that it can locate eye features very accurately. The strength of the proposed method stems from its reproducibility due to the utilization of simple and efficient image processing methods while achieving remarkable results without any need of training.
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-00782224
Contributor : Nesli Erdogmus <>
Submitted on : Tuesday, January 29, 2013 - 12:55:21 PM
Last modification on : Tuesday, January 29, 2013 - 12:55:21 PM

Identifiers

  • HAL Id : hal-00782224, version 1

Collections

Citation

Nesli Erdogmus, Jean-Luc Dugelay. An efficient iris and eye corners extraction method. Joint IAPR International Workshop on Structural, Syntactic, and Statistical Pattern Recognition, SSPR 2010, Aug 2010, Cesme, Turkey. pp.1. ⟨hal-00782224⟩

Share

Metrics

Record views

62