Perceptual Tolerance to Stereoscopic 3D Image Distortion

Abstract

An intriguing aspect of picture perception is the viewer's tolerance to variation in viewing position, perspective, and display size. These factors are also present in stereoscopic media, where there are additional parameters associated with the camera arrangement (e.g., separation, orientation). The predicted amount of depth from disparity can be obtained trigonometrically; however, perceived depth in complex scenes often differs from geometric predictions based on binocular disparity alone. To evaluate the extent and the cause of deviations from geometric predictions of depth from disparity in naturalistic scenes, we recorded stereoscopic footage of an indoor scene with a range of camera separations (camera interaxial (IA) ranged from 3 to 95 mm) and displayed them on a range of screen sizes. In a series of experiments participants estimated 3D distances in the scene relative to a reference scene, compared depth between shots with different parameters, or reproduced the depth between pairs of objects in the scene using reaching or blind walking. The effects of IA and screen size were consistently and markedly smaller than predicted from the binocular viewing geometry, suggesting that observers are able to compensate for the predicted distortions. We conclude that the presence of multiple realistic monocular depth cues drives normalization of perceived depth from binocular disparity. It is not clear to what extent these differences are due to cognitive as opposed to perceptual factors. However, it is notable that these normalization processes are not task specific; they are evident in both perception-and actionoriented tasks.

Vection strength increases with simulated eye-separation

“…Contrary to the proposal that false virtual eye-separations might prevent vection, intriguing perceptual breakdowns (like those described above) appear to be relatively uncommon (e.g., 3-D movie goers typically do not notice reversed disparities when accidental pseudostereopsis occurs at the theatre - Devernay & Beardsley, 2010;Zone, 2005). For the most part, our visual depth, shape, and scene perception appears to be remarkably robust and flexible to conflicts between binocular and monocular sources of information (Allison & Wilcox, 2015;Matthews, Hill, & Palmisano, 2011Ware, 1995;Ware et al, 1998). These findings suggest that visual self-motion perception should be quite tolerant to these types of conflicts as well.…”

Paper: Expert Viewers’ Preferences for Higher Frame Rate 3D Film

“…Rather we set the stereoscopic depth budget for aesthetic reasons, within established norms for S3D filmmaking. Viewers have surprising tolerance to manipulation of stereoscopic viewing parameters 15 but frame rate is expected to influence stereoscopic depth in frame sequential stereoscopic presentations. Our expert viewers preferred HFR for the quality of depth measure which is consistent with theoretical predictions 16 and previous empirical work.…”

16‐2: Invited Paper: Causes and Consequences of IPD Mismatch in XR Devices

“…While the main types of distortion can be broadly categorized as those of size, distance and relative depth, they are closely interrelated due to size and depth constancy. As illustrated in Figure 2 (A), when the rendering camera IA is larger than the viewer's IPD, binocular disparity in the scene is magnified and there is a corresponding increase in perceived depth; a phenomenon known as telestereopsis (30,(33)(34)(35)(36)(37). This expansion of depth causes points in front of the zero-parallax plane (located at infinity for parallel cameras) to appear closer to the observer as indicated by the open circle.…”