Purchase this article with an account.
Zhaoping Li; V1 mechanisms explain filling-in phenomena in texture perception and visual search. Journal of Vision 2004;4(8):689. doi: 10.1167/4.8.689.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Filling-in in texture perception is the phenomenon that a missing or incomplete input texture element in a visual texture is not noticed. A related observation is that it is more difficult to search for a target (termed a minimum target) which lacks a feature present in distractors than to search for a target possessing an unique feature lacking in the distractors (a feature search task). It has been proposed (Li, Neural Computation 13/3. p1749–80, 2001) that a possible explanation for filling-in is the low saliency values of the visual elements near the missing texture element or the minimum target. According to the hypothesis that V1 provides a bottom-up saliency map (Li, TICS, 6, p9–15, 2002), a missing input element, which most likely evokes zero or relatively low visual responses, does not have a high saliency value, and hence it is not easily noticed unless its neighbours (henceforth called the neighbours) are salient enough to attract visual attention to its neighborhood. I apply a biologically-based V1 model to various texture and visual search stimuli to demonstrate the correlation between the low saliencies of the neighbours and filling-in in texture perception or difficulties in search for minimum targets. In particular, I will demonstrate how the degree of textural filling-in or the difficulty of minimum target searches can be altered by altering the saliencies of the neighbours. The saliencies of the neighbours are altered by altering their contrasts or by changing aspects of the stimuli such as features, spatial configurations of the visual elements, and their homogeneities. Furthermore, I will present the results from psychophysical experiments on visual search for minimum targets (and control experiments on non-minimum targets) which test the model predictions.
This PDF is available to Subscribers Only