An efficient iris and eye corners extraction method

Erdogmus, Nesli; Dugelay, Jean-Luc
SSPR and SPR 2010, Structural, Syntactic, and Statistical Pattern Recognition - Joint IAPR International Workshop, Cesme, Izmir, Turkey, August 18-20, 2010 / Also published in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol 6218, 2010

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.

DOI
HAL
Type:
Conference
City:
Cesme
Date:
2010-08-18
Department:
Digital Security
Eurecom Ref:
3254
Copyright:
© Springer. Personal use of this material is permitted. The definitive version of this paper was published in SSPR and SPR 2010, Structural, Syntactic, and Statistical Pattern Recognition - Joint IAPR International Workshop, Cesme, Izmir, Turkey, August 18-20, 2010 / Also published in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol 6218, 2010 and is available at : http://dx.doi.org/10.1007/978-3-642-14980-1_54

PERMALINK : https://www.eurecom.fr/publication/3254