Natural Perspective: Mapping Visual Space with Art and Science

Abstract

Following its discovery in fifteenth-century Italy, linear perspective has often been hailed as the most accurate method of projecting three-dimensional visual space onto a two-dimensional picture plane. However, when we survey the history of European art it is evident that few artists fully complied with its mathematical rules, despite many of them being rigorously trained in its procedures. In this paper, we will consider how artists have actually depicted visual space, and present evidence that images created according to a "natural" perspective (NP) used by artists are judged as better representations of visual space than those created using standard linear (LP) and curvilinear fisheye (FP) projective geometries. In this study, we built a real three-dimensional scene and produced photographs of the scene in three different perspectives (NP, LP and FP). An online experiment in which we asked people to rank the perspectives in order of preference showed a clear preference for NP compared to the FP and LP. In a second experiment, participants were asked to view the real scene and rate each perspective on a range of psychological variables. Results showed that NP was the most preferred and the most effective in depicting the physical space naturally. We discuss the implications of these results and the advantages and limitations of our approach for studying the global metric and geometrical structure of visual space.

Craniofacial superimposition: a review of focus distance estimation methods and an extension to profile view photographs

“…Perspective has long been recognized as an important factor for characterizing 3D objects on 2D planes, not just in photography, but also in architectural illustrations where it provides depth and realism. This has been the case ever since the principles of perspective were first discovered in the Renaissance [ 54 ], and its relevance for, and accountability in, craniofacial superimposition is long overdue.…”

The Art and Science of Displaying Visual Space

“…We constructed various novel apparatus that we used to measure the perceived changes in size and shape of objects across the full visual field of participants (Figure 2). We then attempted to map this structure graphically by applying 2D image deformations to a linear perspective image to produce images that approximated the structure of visual space [41]. We were able to experimentally verify, albeit in limited cases, that these perceptually modelled images were indeed judged as more natural compared to a range of standard projection methods, such as linear perspective, stereographic, fisheye and others.…”

ZoomShop: Depth‐Aware Editing of Photographic Composition

“…Finally, our work is motivated in part by an intriguing observation from the perception literature: in some cases linear perspective photographs may not capture subjective visual experience as well as nonlinear images, particularly those that inflate the size of objects in or near the fovea [BBP14; BPR18; Rau82; KDPP16]. Allowing users to resize objects in images could be useful for better conveying scenes and describing visual experiences [Mat21].…”