Integration of elementary features in texture perception
Final Report Abstract
A long-standing problem in human vision research is how the brain decomposes a visual scene into features and reintegrates them into unambiguous and stable object representations. When objects differ from the surround in several features, they can become more salient than expected from each individual feature. This is known as the "feature synergy" effect. In this project, we investigated the functional role and hypothetical source of feature synergy in the detection and discrimination of 2D texture shapes by (i) combining psychophysical data and modelling of early feature-selective processing, and (ii) recording EEG during simultaneous psychophysical tasks. For textures defined by local modulation of orientation and spatial frequency, we found strong summation of both features in detection and localisation tasks, suggesting integration by a common mechanism. For different texture modulations (local variability and local feature difference), the summation effect could be explained by local energy-based multiscale models, suggesting feature integration at early processing stages. Psychophysical experiments comparing texture target detection and shape discrimination with highly similar targets showed a decrease of cue summation in difficult shape discriminations compared to simpler ones, indicating only a small cue combination advantage for shape judgements after figure-ground segregation has settled. Psychophysical results thus support that feature synergy is limited to detection and discrimination at the figureground segregation threshold where more complex shape processing is not yet feasible. While psychophysical and modelling results clearly point to an early integration mechanism, EEG results revealed a strong reduction in P2 and P3 amplitude as a stable correlate of feature synergy, observed in a cluster of 13 adjacent electrodes from left, central and right occipital and central parietal-occipital lobes. Modulations of P2 and especially P3 amplitudes depended on the psychophysical task (detection vs. shape discrimination) and were not strongly related to perceptual saliency. These findings suggest that P2 and P3 markers of feature synergy originate from regions involved in feature and shape processing under attentional control, reflecting reduced resource allocation when the target is redundantly defined by two features. Notably, no target/non-target modulation was observed in earlier waveforms than P2 for the lower and middle ranges of target-background feature differences used. Thus, our findings do not rule out the presence of earlier EEG correlates of feature synergy and saliency, particularly when stronger feature differences – capable of eliciting segregation-specific responses (tsVEPs, Bach and Meigen, 1992) – are employed.
Publications
-
Feature synergy is based on regional saliency during figure-ground segregation [Conference presentation abstract]. 45th European Conference on Visual Perception (ECVP), Paphos, Cyprus
Hunt, C. & Meinhardt, G.
-
Beneath the Surface: Feature Synergy Improves Texture Segregation but not Shape Perception [Conference presentation abstract]. 46th European Conference on Visual Perception (ECVP), Aberdeen, Scoland
Hunt, C. & Meinhardt, G.
-
The Visual Perception of Figures in Textures: An EEG Source Localization Study [Conference presentation]. 49th Annual Conference Psychologie und Gehirn, Hamburg, Germany
Löffler, C., Hunt, C., Hülsemann, M., Schubert, A.-L. & Meinhardt, G.
-
Electrophysiological correlates of saliency and feature synergy. Center for Open Science.
Hunt, Cordula; Löffler, Christoph; Hülsemann, Mareike; Schubert, Anna-Lena & Meinhardt, Günter
-
Feature synergy enhances saliency of figure-ground segregation. Center for Open Science.
Hunt, Cordula; Schubert, Anna-Lena & Meinhardt, Günter
