A Precise Eye Localization Method Based on Ratio Local Binary Pattern

Abstract

A novel method for human eye location is proposed in this paper, which is designed based on an improved algorithm called ratio local binary pattern (RLBP). These patterns provide a simple but powerful spatial description of eye texture and have low cost of calculation. The working principle of RLBP is described in detail, and based on it a system is developed. The eye location system is composed of three parts namely face detection, eye windows detection and precise eye location. In the face detection stage, AdaBoost method is used to search for the pure face region and crop it from the facial image. Then the RLBP encoding and gray projection are conducted on the pure face and two eye windows are got. After these steps, the boundary tracking and centriod method are adopt for the precise eye location, all the eyes located are annotated and drawn a rectangular box around the center. The experiment has been carried out on a wide set of images extracted from databases, and it shows that the algorithm of eye location is efficient and robust.