It is presently technically impossible to supply technically all the information that can possibly be collected by the human visual system. So in creating virtual realities, it is important to concentrate on improving those parameters of the display that are readily detected by the visual system and to neglect those physical imperfections of the display the visual system is unable to detect. In consequence, the relevant parameters of the display should be optimised as far as possible.
Moreover, one may use conditions that deteriorate the input characteristics thus assimilating virtual to real reality. Under these deteriorated conditions, even an imperfect simulation of the world will give an acceptable feeling of presence since the visual system cannot detect the imperfections present in the display. One contribution of this project was to generally advise how the necessary imperfections of virtual reality may be hidden and to perform quantitative psychophysical experiments on some parameters of the simulation that may be critical for the feeling of presence.
The project started out with the question of how observers experience non-perspective distortions as introduced by a new technique for creating photo realistic 'virtual realities', the two-slit approach which lies at the heart of the BENOGO project. In this project, image-based rendering presentations (IBRs) are the central issue, and psychophysical techniques are used to measure the influence exerted by projections inherently linked to this technique. In addition, two important parameters: luminance and contrast that have a possible influence on the impression of Presence have been investigated quantitatively.
The initial aim was to evaluate to what extent, under what circumstances, and, primarily, due to which cues, IBR images created on the basis of the two slit approach differ from traditional perspective images and how these factors may hinder the sense of 'Presence'. More specifically, we measured the influence of geometrical distortions of three-dimensional reconstructions caused by the two-slit technique as well as the influence of variations in both luminance and contrast.
Over the course of the project, three different lines of experimentation were pursued. The first was to study the sensitivity of the human visual system in detecting distortions for both stationary and moving objects by using well-defined (and hence relatively simple) stationary and moving visual stimuli of the type typically used in psychophysical experiments. It turned out that moving stimuli allow greater distortions and hence projection discrepancies than stationary ones, an insight that could be used in scenario design. The second aim was to apply these (and other) results to make use of imperfections in human vision when creating virtual realities.
The results of these investigations are presented in detail in Deliverable 7.1. The third, most complex approach was to investigate the effect of non-perspective imaging in spatial visualisations to clarify the amount of perspective distortion tolerated for natural stimuli, using complex photo realistic images. The purpose of this study was to inform the demonstrator design process concerning necessary and relevant imperfections of the human visual system, and on how these can be exploited to hide imperfections in the visual renderings of the project's technology as well as to determine the limits of distortion and hence of spatial extrapolation tolerated by different observers. New concepts had to be developed for psychophysical experiments serving this purpose.
Since the first approach lays the foundations of the other two, we started with this project, measuring detection thresholds of human observers for distortions in both straight and curved line stimuli. The results were presented in Deliverable 7.1. The present deliverable will therefore concentrate on the results obtained towards achieving the third aim, measuring the sensitivity of observers for geometrical distortions as well as for variations of both image luminance and contrast. To be able to perform sensible psychophysical measurements with ill-defined stimuli such as natural scenes, a new psychophysical approach was developed for measuring the quality of IBRs, as was mentioned above.
The results show a relatively high tolerance of observers for geometrical distortions, with a significant effect of perceptual learning in the sense that prior experience exerts a certain influence on stimulus preference, and an unexpectedly large amount of inter-observer variance regarding the optimal level of contrast and luminance. This inter-individual variation between observers suggests that individual observers should be enabled to chose the luminance and contrast levels they prefer. This simple adjustment may considerably increase the subjective impression of Presence.