Thesis defence : Alexia ROUX-SIBILON

We understand most of our visual environment at a glance, regardless of its complexity. Theories of visual recognition postulate that this ability is based on the extraction of the gist of the visual scene, a first global and coarse visual representation. Gist perception would be based on the rapid analysis of low spatial frequencies in the visual signal, and would allow a coarse categorization of the scene and the objects that compose it. This first representation of the scene would also trigger predictive mechanisms that guide a more detailed visual analysis. This theoretical framework, which we formalized in this thesis under the name of predictive frequency model, accounts for a large number of empirical results. Overall, work in support of this model has focused on the visual recognition of small stimuli presented in central vision. However, visual acuity decreases drastically with retinal eccentricity so that a large part of the signal at low spatial frequencies comes from peripheral vision. In this doctoral thesis, we tested the hypothesis that the low spatial resolution information extracted in peripheral vision would influence the processing of detailed information in central vision. In a first series of experiments, we observed that the categorization of complex visual stimuli (objects or scenes) presented in central vision can be improved and / or accelerated when information in peripheral vision is semantically congruent compared to when it is inconguent. In a second series of experiments, we studied the consequences of a peripheral vision loss onto visual recognition in central vision. We tested patients with glaucoma. This ophthalmological condition affects the peripheral retina, thus representing a good pathological model of a system in which peripheral vision is degraded. We observed that these patients presented a deficit to categorize small scenes presented in central vision, even though their retinal sensitivity was preserved for this region of the visual field. This last result suggests that the loss of peripheral vision disrupts the predictive visual system and would thus disturb visual recognition globally, even in central vision. Overall, the results allow consolidating and clarifying the predictive frequency model while emphasizing the relevance of considering peripheral vision in visual recognition models

Encadrante :
- Directrice de thèse : Carole PEYRIN - carole.peyrinuniv-grenoble-alpes.fr (carole[dot]peyrin[at]univ-grenoble-alpes[dot]fr)
Keywords: Glaucoma, Scene perception, Predictive coding, Peripheral vision

Read the thesis