Comparison of Ear Recognition Robustness of Single-View-Based Images Rotated in Depth

To improve the robustness against variation in shooting angles, we previously proposed using an asymptotic expansion of the Gabor transform of ear images to compute the Gabor features of other poses and using these estimates in multiple linear discriminant analysis to enhance feature discriminability. Extending this study, the accuracies are compared with other standard methods that can be used to compute feature vectors for other poses, as in principal component analysis and multiple regression analysis.