Stereopsis is the process of creating a three-dimensional (3D) illusion based on two slightly different projections of a scene. Each projection is projected onto one eye. This effect, exploited in photographic images since the 19th century and, more recently, in an abundance of 3D movies, exists because of the eye’s horizontal separation.

Wijntjes and Pont’s paper is a study of the differences in depth perception based on the use of three natural scenes: a bird flock, a flower bouquet, and a scene of a group of people. Mono and stereo images of these scenes are presented to a group of eight subjects in random order. In each session of the experiment, subjects are asked to perform a pointing task--adjust a pointer, manipulated by a computer mouse, to point toward an object in the images. The stimulus (a dot on the image) is only presented to the left eye, in order to avoid disparities when the stereo images are presented.

As the data is limited and the natural scenes are fairly complex, not many firm results are reported. The authors discuss observations and rationalize why some of the observed parameters may be surprising. One of the conclusions is that for two of the three scenes, the relative depth perception is better in stereo than in mono. The subjects also indicate that the stereo images have more depth. However, it seems that the improvement from binocular disparity typically depends on the contents of the scene itself.

Overall, this is an interesting paper for those studying cognitive phenomena in humans, or those trying to build efficient, anthropomorphic human-computer interfaces and virtual environments.