Ear recognition after ear lobe surgery: A preliminary study

Abstract

Earlobe surgeries are performed with the intention to correct the ear characteristics both locally and globally and also to beautify the appearance. Since performing the surgery (both for beautification and corrections) will alter the original ear features to the greater extent thereby poses a significant challenge for ear recognition. In this work, we introduce and explore this problem of ear recognition after ear lobe surgery. To this extent, we prepared a new ear surgery database comprising of 50 subjects with both pre and post surgery ear samples. We then propose a new scheme for ear recognition based on the hybrid fusion of block features extracted from the ear images using Histogram of Oriented Gradients (HoG) and Local Phase Quantisation (LPQ). We present extensive experiments on the ear surgery database by comparing the performance of eight different state-of-the-art schemes to study the effect of ear surgeries on ear recognition accuracy. The results on the ear surgery database indicate a great challenge as the eight different state-of-the-art schemes are unable to provide acceptable levels of identification performance.