Fast switching of cue integration weights O Watanabe1, M Matsuda2, R Tamura2 |
|---|
1College of Information and Systems, Muroran Institute of Technology, Japan |
| To perceive three-dimensional (3D) structures of external environments, the visual system uses multiple depth cues such as binocular disparity and texture gradient. These cues are extracted in individual modules and then integrated. An optimal way to integrate multiple cues is the maximum likelihood estimation (MLE) [Ernst and Bulthoff, 2004, Trends in Cognitive Sciences, 8, 162-169]. The MLE provides the most precise depth estimate by weighting cues according to those reliabilities. Although various experiments indicate the brain employs the MLE to integrate multiple module/modality information, some open questions remain. Here we examined whether the brain can change the integration weights immediately after we see a new scene. Cue reliability depends on scenes and objects as well as the nature of the cue itself. In the experiment, three slanted natural images with disparity and texture cues were presented in random order. Observers had a prior knowledge concerning the images, and, if the visual system could change the weights rapidly, they should perceive the image slants with minimal variance in every trial. The result showed that observer’s judgments were as precise as the MLE and suggests that the visual system switched the weights appropriately according to what image was presented in each trial. |
| Up Home |
|---|