"Neural mechanisms supporting efficient natural scene processing "
Our everyday visual environments contain a large amount of objects. Due to the limited capacity of the visual system, these objects need to compete for neural representation. Despite this competition, humans can perform strikingly well in complex real-world environments: for example, the detection of object categories in cluttered natural scenes is remarkably rapid. I will highlight two brain mechanisms that support this efficiency. First, I will discuss how attention modulates neural category processing to facilitate selection in natural scenes. Using multivariate decoding of MEG data, we have demonstrated that attention quickly resolves competition between objects, allowing for the rapid neural representation of task-relevant objects. During a category detection task in natural scenes, MEG response patterns within the first 200ms of processing only reliably carried category information about behaviorally relevant targets (but not distracters), suggesting that attention functions as an effective filter at the stage of perceptual category processing. Second, I will discuss how real-world object regularities are used by the brain to support efficient scene parsing. We have shown that pairs of objects can be grouped together when they are arranged in a regular configuration that is commonly seen (e.g., lamp above table), but not when arranged in an irregular one (e.g., lamp below table), thereby reducing scene complexity. This grouping process enhanced category-selective processing in category-selective areas of visual cortex (fMRI), and in early category-selective waveforms (MEG). Importantly, inter-object grouping also facilitates target detection during search, indicating that it is capable of supporting target detection in complex natural environments.
May 02, 2016 | 04:00 PM - 06:00 PM