IMPORTANT : L'inscription est gratuite mais obligatoire.
Les GDR MSPC et Vision organisent une journée conjointe à l'interface des mathématiques, de la vision artificielle et de la vision biologique.
La journée sera ponctuée par quatres exposés donnés par des experts de ces domaines. Elle sera également l'occasion de discussions informelles.
Date
Mardi 17 Novembre 2009
Localisation
Institut Henri Poincaré (Paris, France)
Amphitheatre Hermite
Comité d'organisation
Programme.
10:00-11:00 - James Elder (York University)
"Segmenting Salient Shapes"11:00-12:00 - Agnès Désolneux (Paris 5)
"Le principe de Helmholtz pour la perception visuelle"14:00-15:00 - Pascal Mamassian (Paris 5)
"Visual perception under uncertainty"15:00-16:00 - Yves Fregnac (UNIC, Gif-sur-Yvettes)
"Multiscale functional imaging of the dynamic "association field" in V1"
Résumés.
James Elder - "Segmenting Salient Shapes"
I will discuss the problem of detecting and segmenting salient objects in natural images. Humans are very good at this, and very fast. Most thoroughly studied is the task of rapid animal detection, however the mechanisms underlying rapid detection remain poorly understood. Here I will report recent psychophysical results suggesting that the fastest mechanisms underlying animal detection in natural scenes use contour shape as a principal discriminative cue, while somewhat slower mechanisms integrate these rapidly computed shape cues with image texture cues. Detection continues to improve with increased stimulus exposure, suggesting progressive refinement of neural representations. These results pose a challenge for computational vision algorithms, as the performance of current contour grouping algorithms falls short of human perception. Here I will present results of a Bayesian coarse-to-fine contour grouping algorithm in which approximate representations are computed rapidly and then refined over time. The algorithm outperforms single-scale algorithms and suggests one possible role for massive feedback projections from higher cortical areas in the object pathway to earlier visual areas.Agnès Désolneux - "Le principe de Helmholtz pour la perception visuelle"
D'après le principe de Helmholtz, toute structure qui présente trop de régularité géométrique pour être présente par hasard dans du bruit attire notre attention et devient alors un "objet visuel". Dans cet exposé, j'expliquerai comment ce principe, combiné aux principes de groupement visuel de la théorie de la Gestalt, peut-être utilisé pour calculer effectivement des seuils automatiques de détection.Pascal Mamassian - "Visual perception under uncertainty"
Visual perception can often be seen as an inference problem where the available retinal information is uncertain, its interpretation is ambiguous and perceptual decisions depend on the task of the observer. Uncertainty, ambiguity and task-dependency can be modelled within a Bayesian probabilistic framework with the likelihood, prior and utility functions respectively. Several examples will illustrate how this framework can be used in visual psychophysics.Yves Frégnac - "Multiscale functional imaging of the dynamic "association field" in V1"
In vivo intracellular electrophysiology offers the unique possibility of listening to the "synaptic rumour" of the cortical network captured by the recording electrode in a single V1 cell. The analysis of synaptic echoes evoked during sensory processing is used to reconstruct the distribution of input sources in visual space and in time. It allows to infer, in cortical space, the dynamics of the effective input network afferent to the recorded cell. We have applied this method to demonstrate the propagation of visually evoked activity through lateral (and possibly feedback) connectivity in the primary cortex of higher mammals (Frégnac and Bringuier, 1996; Bringuier et al, 1999; Series et al, 2003). The predictions of macroscopic network behaviour, based on microscopic functional synaptic imaging, have been recently validated by real-time functional network imaging technique, based on the use of voltage-sensitive fluorescent dyes (Jancke et al, 2004; Roland et al, 2006; Benucci et al, 2007; Xu et al, 2007; Nauhaus et al, 2009). The joint application of the two techniques which address two different scales of integration is used to elucidate the cortical origin of low-level (non-attentive) binding processes participating in the emergence of illusory motion percepts predicted by the psychological Gestalt theory (Chavane et al, submitted; Frégnac et al, 2009).
We will review two cases of motion illusion (apparent motion and line motion), which suggests the genesis of a wave of perceptual binding operating in V1, which modulates the integration of feed-forward inputs yet to come: this wave can be seen as the propagation of the V1 network belief of the possible presence of a global percept (the “whole”:) before the illusory percept becomes validated by the sequential presentation of the “parts” (signalled by direct focal feed-forward waves). This neuronal dynamics obeys closely the Gestalt prediction that the emergence of the “whole” should precede in time the detection of the “parts”.
We propose that the primary visual cortex plays the role of a generalized echo chamber fed by other cortical areas (visual or not) which participate in the coding of shape and motion in space: accordingly, the waves travelling across V1 would signal the emergence of perceptual coherence when a synergy is reached between the different cortical analyzers.
