抄録
Numerical simulations were performed using super-resolution images and barrel-distorted images to improve the viability of the authors' previous work regarding single-view-based ear biometrics. Empirical demonstrations show that, using barrel-distorted images and super-resolution images, accuracy decreases. However, our present proposed method for model normal vector creation applied to these images compensated for losses in accuracy. Furthermore, an experimental investigation of the influence of earrings on results is included. The difference between the averaged rank between images of subjects with and without earrings at each input angle was found to be statistically insignificant.
