Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.567208
Title: Norm-­ and exemplar-­ based models of face recognition
Author: Ross, David Andrew
Awarding Body: Cardiff University
Current Institution: Cardiff University
Date of Award: 2011
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
Face recognition is vital for many of our day-­‐to-­‐day social interactions. For the most part we can effortlessly recognise the faces of family, friends, casual acquaintances and famous people, ignoring the new hairstyle or fashionable sunglasses that they might be sporting and attending to diagnostic features that reveal their identity. From a psychological or computational standpoint, our flair for face recognition is particularly interesting because, unlike the many other categories of object that we routinely encounter, faces must be individuated. That is to say, faces must be differentiated at the within-­‐category level, placing a unique demand on the visual system’s ability to rapidly and accurately discriminate a large number of visually similar patterns. The ease with which we recognise familiar faces belies the computational complexity of the task, rendering us unaware of the dramatic image variance caused by changes in viewing angle, lighting conditions and partial occlusion, and leaving us with an illusion of stability that is characteristic of our impression of the visual world. Clearly, with some 120 million retinal cells semi-­‐ independently encoding the various aspects of a visual scene (Palmeri & Cottrell, 2010), the problem of face recognition is not (normally) one of visual acuity. Rather, the problem of face recognition is one of dimensional reduction, representing faces in a way that ignores the gross image-­‐level variance across different instances of the same person while still maintaining the ability to discriminate between different faces (see Figure 1.1)
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.567208  DOI: Not available
Keywords: BF Psychology
Share: