Sampling Static Perspective Cues

This paper primarily deals with sampling artefacts in interactive, immersive 3D CGI. However, before we can consider the perception of sampled 3D images in these dynamic situations, we need to analyse the presentation and perception of sampling in static images. First, we use the static case to establish some of the assumptions and conventions we use in this thesis to describe perspective geometry. Second, we identify and discuss the effects of spatial sampling artefacts in the presentation of perspective depth. Then we describe experiments performed to identify the visual contexts in which these artefacts are likely to impair the judgement of depth. Finally, we describe and analyse two effective methods for ameliorating these artefacts. Given a vanishing point located at the centre of the screen and the number of pixels, (nh, nv), we can compute the exact location of the 3D point in 2D screen coordinates: xs = (x ex) nh e zys = (y ey) nv ez sh ez z sv ez z To get the sampled location of a point, we round (xs, ys) to the centre of the nearest pixel. In the equations above, we assume the line of sight orthogonally intersects the centre of the screen, a useful and frequently used convention. However, the location and orientation of the viewpoint plays a critical role in the perception of scene layout. It is impossible for a user to discriminate two points separated along the line of sight because the points share the same location on the screen. However, if the viewpoint is rotated around the two objects so their difference lies perpendicular to the viewers line of sight, discriminating the two points is substantially easier.