Exploring Ear Recognition: Image Fusion and Deep Learning with Thermal and Visible Images

Abstract

This pioneering study delves into the realm of ear recognition, introducing an innovative methodology that combines thermal and visible ear images using sophisticated multiresolution analysis, including discrete wavelet, ridgelet, and curvelet transforms. Our tailored deep learning model, uniquely designed for ear recognition, showcases outstanding results, with the complex-valued curvelet transform paired with thermal images achieving an impressive 96.82% recognition rate—outperforming all other methods. Notably, this research underscores the transformative impact of multi-source data fusion on elevating the effectiveness of ear recognition systems. Key Words: ridgelet, curvelet, effectiveness, multiresolution, pioneering