Eyelash Detection Based on Coefficient of Variation and Gradient Weighted Direction Filtering

Aiming at the problems of current eyelash miss detection and balancing detection accuracy and speed difficultly, this paper proposes an eyelash detection algorithm based on coefficient of variation and gradient weighted direction filtering. Firstly, the coefficient of variation criterion is designed to determine the eyelash occlusion region, and then the minimum intra-class coefficient of variation method is used to complete the eyelash root detection. Secondly, the multi-scale composite window and the gradient vector weighted projection are combined to determine the tail eyelash direction. Finally, the dynamic direction filter is used to detect the low-contrast and multi-directional tail eyelash. On the CASIA-IrisV1 and CASIA- IrisV3-Interval databases, compared with the traditional detection algorithm based on Gabor filtering and regional gray variance detection, the detection algorithm based on eyelid contour and local gray minimum, and the detection algorithm based on morphological operation, the experimental results show that the proposed algorithm is superior to other compared algorithms in subjective accuracy (coincidence between detection results and manual marking results), detection time (algorithm complexity analysis), eyelash missed detection rate (false eyelash-detection rate, FER) and eyelash error detection rate(false non-eyelash-detection rate, FNER), and has strong robustness. © 2020, Beijing China Science Journal Publishing Co. Ltd. All right reserved.