Purchase this article with an account.
Jeremiah D. Still, Veronica J. Dark, Derrick J. Parkhurst; Viewpoint invariant object features attract overt visual attention. Journal of Vision 2007;7(9):445. doi: 10.1167/7.9.445.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Previous research suggests that salient image features attract overt shifts of attention when participants freely view complex artificial and natural scenes (Parkhurst, Law & Niebur, Vision Research, 2002). This evidence is consistent with the influence of a bottom-up stimulus-driven mechanism of attentional guidance. Attention to salient image features is a plausible default information selection strategy, especially in the absence of a well-defined task.
Given that object recognition is necessary for most natural visual tasks, a plausible alternative default strategy is to preferentially select information likely to be important for object recognition. Object recognition depends in part on the presence of visual features that remain invariant across viewpoints (Biederman, Psychological Review, 1987). Thus, it is possible that viewpoint invariant features of objects will attract visual attention.
To examine this possibility, the eye movements made by 12 participants freely viewing images of objects were recorded. Within each image, the Scale Invariant Feature Transform (SIFT) algorithm was used to identify highly invariant features (Lowe, International Conference on Computer Vision, 1999) and the Saliency model was used to identifiy salient features (Itti, Koch & Niebur, IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998). Because the predictions of the two models can be highly correlated, we selected images of 100 different objects from the Amsterdam Library of Object Images (Geusebroek, Burghouts & Smeulders, International Journal of Computer Vision, 2005) that maximally decorrelated the predictions.
Both models performed significantly better than chance. The SIFT model performed signifcantly better than the Saliency model. These results suggest viewpoint invariant features of objects attract attention as reflected in eye movements. They also support the hypothesis that the default attentional selection strategy is biased to select visual features likely to be important for object recognition.
This PDF is available to Subscribers Only