Eye localization for face matching: is it always useful and under what conditions?

In this paper we address the influence of eye localization accuracy on face matching performance in the case of low resolution image and video content. By means of a broad experimental evaluation involving several base-line eye localizers and face matching algorithms we investigated to which extent and under what conditions the eye localization accuracy will benefit the face matching performance, both in terms of effectiveness and efficiency. Our evaluation showed that (1) worst-case eye localization errors have a big impact on face matching performance, (2) in respect to that and under a minimum required accuracy, eye localization can boost performance of naive face matchers, (3) eye localization allows for more efficient face matching without degrading performance.