University of Minnesota
Object Segmentation in Complex Natural Scenes Modulates fMRI Early Visual Cortex Responses
Abstract
dc:description.abstractVisual perception requires organizing local image features into coherent objects and background regions. Using 7T fMRI data from the Natural Scenes Dataset and COCO object annotations, this dissertation tests whether object segmentation is reflected in human early visual cortex during natural scene viewing. Neural populations whose receptive fields fell on segmented objects showed stronger responses than those responding to background regions. This enhancement was reliable across participants and visual areas, and persisted after accounting for low-level feature predictions, segment size, salience, and the number of objects in a scene. Segmentation also shaped trial-to-trial covariability. Across repeated image presentations, neural populations showed stronger covariability when they represented locations within the same segment than when they represented different segments. Contextual modulation models provided a potential link between these effects, suggesting that objecthood changes contextual gain while segment membership changes contextual interactions across feature channels. Importantly, although the model was trained only to predict mean responses, adding feature-level noise produced same-segment covariability as an emergent effect. These findings show that human early visual cortex carries segmentation-related information during natural scene viewing, reflected in both response magnitude and population covariability.
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Nie, Shaozhi
Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Citation
- Nie, Shaozhi. (2026). Object Segmentation in Complex Natural Scenes Modulates fMRI Early Visual Cortex Responses. Retrieved from the University Digital Conservancy, https://hdl.handle.net/11299/283133.