Kernel graph cut for robust ear segmentation in various illuminations conditions

Ear as biometrics is still in its infant stage. Further research can be explored using ear as recognition of a subject and one of the vital stages to be investigated is segmentation of the ear. Hence, in this paper we address an enhanced technique of ear segmentation that has improved recognition accuracy as compared to our previous method. The proposed method adapted the Kernel Graph approach and succeeded in performing segmentation under various illumination conditions. Initial findings showed that the proposed method attained 100% accuracy as compared to our earlier technique with 95% accuracy rate only.