アイテム詳細

要旨

Face recognition relies on detecting subtle differences between facial parts (components) and their spatial relations, so-called configural information (for a recent review see [2]). Psychophysical experiments are reported, in which a new method was used to test component and configural processing separately [3]. An important role for both types of information was revealed in unfamiliar as well as familiar face recognition and only quantitative but no qualitative differences were found. The results could be modeled using a computational framework based on key-frames in which the component and configural route were explicitly implemented [4]. In a new series of experiments, the Inter-Extra-Ortho paradigm by B¨ulthoff and Edelman [1] was applied to investigate face recognition across viewpoint. Systematic effects of viewpoint were found that are consistent with local view interpolation schemes and the proposed framework based on key-frames [5]. In a recent study the role of component and configural information in view-based face recognition was investigated by combining the Inter-Extra-Ortho paradigm with the psychophysical method for selective testing of processing component and configural information. Interestingly, processing of parts was more dependent on viewpoint than processing of configural information. The same result was found for the computational model mentioned above, in which the component and configural processing route have been implemented. The psychophysical and computational results are also discussed in respect to a third route to recognition in which different views of a face are temporally associated [6].