Part-based strategies for visual categorisation and object recognition
Approaches to object recognition that rely on structural, or part-based, descriptions have a long-standing tradition both in computer and human vision. Yet despite common roots the synergetic potential of part-based techniques from machine vision for the understanding of human object recognition has remained largely under-utilised. Here I propose one such approach - evidence-based systems (EBS) - as a process model for human learning of pattern or object categories. Its usefulness is demonstrated by analysing behavioural data in experiments that address the effects of context and of mirror-image relations in pattern category learning – aspects that are difficult to assess by traditional psychometric categorization models. Finally, the implications of these experiments and their computational analysis for current theories of object recognition are discussed. It is argued that generic part-based recognition strategies, as exemplified by EBS, provide a promising way to reconcile divergent positions in the debate between proponents of structural models on the one hand and image-based models on the other.
Jüttner, M. (2007) Part-based strategies for visual categorisation and object recognition. In: N. Osaka, I. Rentschler & I. Biederman (eds.) Object Recognition, Attention & Action (pp. 55-70). Springer, Tokyo.