A Novel Eye Localization Method Based on Log-Gabor Transform and Integral Image
- 软件学院－已发表论文 
A robust and precise eye localization method plays a critical role in automatic face recognition system. In this paper, a novel eye localization approach using Log-Gabor transform and integral image is proposed. This method mainly involves three steps: firstly, we adopt the Log-Gabor transform with certain scale and orientation to locate the approximate eye-and-brow region. Secondly, in order to detect the coarse region of each eye, the integral image of the Log-Gabor transform result is calculated. Ultimately, we compute the integral image of the original face image within the coarse regions to implement the fine localization. Experimental results on the FERET database show that the method can precisely locate the eyes on both frontal faces and rotated faces in plane within about twenty-five degrees.