Illumination Invariant Face and Pupil
Detection 1Department
of Applied Computer Science, Tokyo Polytechnic University {yskang, arai, ono}@cs.t-kougei.ac.jp
2Department
of Electronics and Mechatronics, Tokyo Polytechnic University d.shin@em.t-kougei.ac.jp
Abstract Personal video surveillance systems have been developed in the intelligent security systems and monitoring research areas. Computer vision technologies are key to the implementation of surveillance and monitoring systems, applying methods such as face recognition, pupil detection, and human motion analysis. One of the most important techniques is robust illumination invariant object detection for surveillance systems. This paper presents illumination invariant face and pupil detection using a multiband camera for a personal surveillance system. The multiband camera provides images in four bands, near infrared and the three RGB color bands. We introduce a multiband camera system that utilizes two different images, i.e., a color image extracted from a Bayer filter and a near infrared image. We employed a Viola-Jones Face Detector using Haar-like features and AdaBoosting classification for face detection. We also proposed a pupil detection method using a differential of red-band and near-infrared images. Results showed that the proposed multiband camera system is effective for implementing an illumination invariant object detection module. Keywords: Personal
surveillance system, Illumination invariant, Face and pupil detection, +: Corresponding author: Yousun Kang Kanagawa, 243-0297, Japan, Tel:+81-46-242-9524 IT Convergence Practice (INPRA), Vol. 5, No. 2, pp. 1-10, June 2017 [pdf] |