This paper investigates a new approach for reliable personal identification using gray-level ear images. The developed approach extracts robust phase information using 2-D quadrature filtering (both monogenic and quaternionic). The use of quadrature filters is motivated by their ability to jointly localize spatial and frequency domain phase information in the segmented ear images. This paper has developed two new techniques for biometric recognition: the quaternionic quadrature filter and the QuaternionicCode. We comparatively evaluate the performance of 1-D and 2-D local phase information using Cauchy, Gaussian derivative and log-Gabor band-pass quadrature filters. We detail experimental results, both for recognition and verification, using publicly available UND ear dataset and IITD ear database. The achieved results with 2-D quadrature filters are highly promising and achieve significantly improved performance as compared to conventional phase encoding using 1-D quadrature filters employed in the literature.
- Ear recognition
ASJC Scopus subject areas
- Signal Processing
- Computer Vision and Pattern Recognition
- Artificial Intelligence