The two main goals of the Active Vision Laboratory are:
- developing sophisticated rendering systems that will allow the generation of more realistic and ecologically valid displays;
- developing computational models of 3D perceptual processes and testing the biological validity of these models with the rendering tools.
3D information for perception and action
The main goal of the Active Vision Laboratory is to developi experimental paradigms that will allow us to have a more complete control over 3D cues to depth. The aim is to understand how 3D cues influence perceptual and motor judgments.
Motor judgments will be studied by having observers reach-to-grasp virtual or real objects. We will monitor the observers movements, typically the movement of their hand and fingers, through an OPTOTRAK system.
Our most ambitious project is to create the most realistic 3D displays: holograms. Holograms are different from stereograms, since the former generate points of light in 3D space, whereas the latter are pairs of images on flat screens. Therefore holograms are void of cues to flatness. The holographic system will be comprised of a laser combined with a Spatial Light Modulator (SLM). Whereas building such system does not entail serious technical challenges, most of the effort will be involved in programming the SLM to create the desired 3D patterns
Full cue environments and “real objects”
Computer generated displays can be easily manipulated by the experimenter and therefore constitute a useful tool for controlling 3D cues in the 2D rendered images. However, residual and uncontrolled cues are present in these displays, which specify the true flat surface of the monitor.
Real surfaces produce a richer stimulation, which includes less studied but nevertheless effective cues like the blurring gradient, motion parallax, accommodation cues and ocular convergence cues. In this research project we intend to study the influence of these cues on the perception of 3D shape by utilizing real objects of which we can control with high precision the 3D structure.
In collaboration with the Robotics Unit of IIT we built a mechanical device that will control the mutual 3D position of metal rods. This apparatus will be combined with a traditional rendering apparatus and, through the use of half-reflective mirrors, we will be able to superimpose virtual images to the real objects and scenes. This hybrid combination of virtual and real objects will allow us to fully understand the effectiveness of the cues-to-flatness that are commonly present in traditional displays.
When an observer moves, the retinal projections of 3D objects in the world change accordingly. This investigation is aimed at understanding how the sensing of body movements allows human observers to aid the interpretation of visual stimulation. This study will involve a computational analysis, in order to establish the relationship between the dynamic information within the retinal images and the extra retinal information concerning head movements. This computational analysis will yield the development of a computational model, which will be used to predict empirical data.
A prototype of the active vision apparatus utilizes a OPTOTRAK Certus system, which allows the precise localization of infrared markers that are, in turn, rigidly attached to the observers’ head. The localization of the markers allows to precisely track the 3D position of the observers’ head. This information is transmitted in real time to a computer that generates the images of 3D objects on a Hi-Res computer display. The images are generated by a program that calculates the projections of the 3D object on the monitor display, so to produce the correct visual stimulation on the observers’ retinas.