A Comparative Analysis of Two Approaches to Periocular Recognition in Mobile Scenarios

Abstract

In recent years, periocular recognition has become a popular alternative to face and iris recognition in less ideal acquisition scenarios. An interesting example of such scenarios is the usage of mobile devices for recognition purposes. With the growing popularity and easy access to such devices, the development of robust biometric recognition algorithms to work under such conditions finds strong motivation. In the present work we assess the performance of extended versions of two state-ofthe-art periocular recognition algorithms on the publicly available CSIP database, a recent dataset composed of images acquired under highly unconstrained and multi-sensor mobile scenarios. The achieved results show each algorithm is better fit to tackle different scenarios and applications of the biometric recognition problem.

Multi-source deep transfer learning for cross-sensor biometrics

“…Furthermore, performance observed for all cross-sensor scenarios was consistently worse than their single-sensor counterparts. This idea is also corroborated by the conclusions of Monteiro et al's work [18], where two state-of-the-art recognition algorithms fail to cope with the variations introduced to their input data when faced with cross-sensor scenarios.…”

“…1, whose potential as a biometric trait can be motivated as a representation in between face and iris recognition. The periocular region has been shown to present increased performance when only degraded facial data or low-quality iris images are made available, and even in mobile application scenarios, it does not require rigid capture or complex imaging systems, thereby making it easy to acquire even by an inexperienced user [18]. While Santos et al propose a framework based on multiple descriptors to work on periocular data on multiple mobile sensors, Jilela and Ross attempt to match iris and face images from the same individual, acquired with distinct sensors, using periocular traits to help in the recognition process.…”

Heterogeneity Aware Deep Embedding for Mobile Periocular Recognition

“…Since the proposed method requires training, a direct comparison with [27] is not feasible. Monteiro et al [18] have also computed the results on this dataset, however cross sensor experiments were not performed results presented in Tables 1 and 2 corroborate the effectiveness of the proposed model. Table 3 summarizes the Rank 1 accuracies of the proposed method on the VISOB Database [26] for the experiment (a) (described in section 3.2).…”

“…Since the proposed method requires training, a direct comparison with [27] is not feasible. Monteiro et al [18] have also computed the results on this dataset, however cross sensor experiments were not performed Table 1: Results on the CSIP dataset for cross-sensor mobile periocular recognition tasks.…”