ACM Computing Surveys (CSUR)
| Hi-index | 0.00 |
This paper examines the feature extraction method based on force field transformation and develops a new two-stage approach for multi-pose ear feature extraction and recognition, i.e., force field transformation plus null space based kernel fisher discriminant analysis (NKFDA). Force field transformation assumes the pixel in the ear image as the particle that acts as the source of a force field. This transformation can strengthen the edges of ear image, which is the most important feature for ear to distinguish from one another. After the force field transformation, NKFDA is employed to extract feature for multi-pose ear image. Kernel technique can not only efficiently represent the nonlinear relation of data but also simplify the NLDA. The experimental results show that the proposed method is more robust and effective than the initial feature extraction method based on force field transformation, the potential well-based technique and demonstrate effectiveness for multi-pose ear recognition.