Effective Eye Localization using Local Binary Patterns
S. S. Shylaja, K. N. Balasubramanya Murthy, N. Sudha, Nitin Kumar, Ruby Agarwal · 2012
Eyes are one of the most salient features of the human face, playing a critical role in understanding a person’s desires, needs and emotional states. They are also considered to be non-deformable objects appearing under various poses and lighting conditions. Therefore, efficient eye localization is a necessary step in many face-related applications like face recognition, face registration, face validation, gaze tracking, blink detection and red eye detection. In this paper, a probabilistic eye localization method based on local binary patterns (LBPs) is presented. Local binary pattern generates a binary code that describes the local texture pattern by normalizing the intensity values in a neighborhood. These patterns provide a simple but powerful spatial description of texture, and are robust to the noise typical to various illumination conditions and pose. LBPs are used for their higher accuracy rate and lower complexity. For a given close-up image, the centre of the iris of two eyes is located. The complete system has been tested on the standard databases and web-cam pictures of people under different light conditions. The accuracy has been nearly 98 % (±1 pixel shift).