Agrawala: Manipulating Surface Shading to Enhance Shape and Surface Details
Artists, illustrators, and photographers commonly manipulate shading to emphasize both the shape and the fine-scale surface details of an object. Often the manipulations produce physically inconsistent or impossible shading effects, but nevertheless serve to increase the amount of shape information presented in a single view of the object. In this talk, I will first describe some of the design principles behind these non-photorealistic shape and detail enhancement techniques. I will then present algorithms and interfaces that are based on these principles and allow users to enhance shape and surface details in digital photographs as well as 3d renderings of objects.
Banks: Why Pictures Look Right when Viewed from the Wrong Place (and sometimes look wrong when viewed from the right place)
When a picture is viewed
from positions other than its center of projection, there can be large
changes specified in the retinal image, yet the perceived spatial
layout and shape of objects do not seem to change. We have shown that
compensation for oblique viewing occurs provided that the viewer can
estimate the slant and tilt of the picture surface accurately.
Compensation is nearly veridical with binocular viewing at close range.
Compensation generally does not occur with monocular viewing through a
small aperture; instead, the percept is dictated by the shape of the
retinal image. Our findings help to explain invariance for incorrect
viewing positions, and other phenomena like perceived distortions with
wide fields of view and the anamorphic effect. Our findings also have
relevance to the design of displays. We will discuss, for example, how
the viewer's position ought to affect percepts depending on the shape
of the display surface. We are currently investigating similar
phenomena with stereographic pictures and finding that quite different
rules apply.
Cohen: Capturing and Viewing Big, Wide, and Deep Imagery
I will discuss new modalities for capturing multi-gigapixel (big), panoramic (wide), and high dynamic range (deep) images. And perhaps more importantly, I'll discuss new tools for exploring such images. I say exploring, rather than viewing, because the complete content of the image is never present on the screen at any one time. Such images are in some ways a new media type in that they are static, like an image, but become dynamic, like a video, through user exploration.
Girshick: The effect of viewing position on the perception of stereoscopic and conventional pictures
When viewing conventional pictures, observers have demonstrated a striking
amount of perceptual invariance for their viewing position. That is, the 3D structure of a picture’s contents
is perceived reasonably accurately, even if the observer is at the incorrect viewing
location. Previous work suggests this is
done using local “surface cues” (information about the picture surface slant). Stereoscopic pictures are similar to
conventional pictures in that they are displayed on 2D surfaces with surface
cues. They differ in that the contents
are specified by binocular disparity. I
will discuss whether invariance occurs with stereoscopic pictures and the
implications for picture perception.
Gopnik: Ken Burns' "Baseball" and Vermeer's "View": A Much Closer Look at Dutch Art
Photographs and movies have conditioned us to assume that pictures render scenes
from a viewpoint in the center of the pictorial frame -- that normal pictorial
"looking" is looking straight into the middle of a scene. (When a movie shows a
train that's coming right at us, it's always heading right out of the middle of
the screen.) They've also trained us to imagine that pictures should be viewed
from some fair distance off. (A snapshot held at arm's length is usually being
viewed from many times the viewing distance implied in the scene it depicts.
Ditto for almost any printed reproduction of a perspectival work of art.) A huge
number of paintings from 17th-century Holland, on the other hand, were
constructed with perspectival viewpoints ("centers of projection") that are
highly off-center, and that imply a spectator very near to the depicted scene.
(Almost inside it, in the case of some still lifes.) Dutch art lovers would have
been keenly aware of the implications of a picture's perspectival construction,
and would have known to rove in front of a painting until arriving at the
"ideal" position from which it should be viewed -- which is often very close
indeed, and far off to one side. Looking at Dutch pictures as they would have
been looked at in their time profoundly affects how they appear, and even what
they mean.
Hughes: Looking at Inifinity from the Other End
I am going to show one of my reverspective paintings, of buildings on the Grand Canal in Venice. This is constructed in perspective in reverse, so the notional vanishing points are actually nearer to the spectator than the picture plane. By using a light source, a spotlight or projector, I will show how important shading is in differentiating between planes. When the difference is accentuated, the planes seems to move in concert with the movement of the seer. My experience as an artist in constructing and painting these things is that the accretion of detail is also important: the buildings of Venice are an example. I suspect these perceptions of perspective are built into us, with our knowledge of verticals and horizontals, which are derived from the force of gravity. When verticals and horizontals are seen askance as walls they become trapezia, and it is with trapezia that I always work. I will also discuss the hollow-mask illusion and the camera obscura.
Johnson: The Majestic vs The Real
With film's reality distortion field, we have heaped even more distortion on our already glorified views of the national parts. It is certainly true that a painter like Albert Bierstadt, with all of the beauty of his work, was not painting just the Yosemite he saw, but the Yosemite he felt and then imagined. As a consequence, viewers of his work saw as much a Yosemite of Bierstadt's mind as one they might see through their own eyes. But that's not how the paintings were seen; inevitably they were thought to be depictions of the place. Color photography, and to some degree black and white photography, has led viewers down that same path. The photographer's choice of the rare moment, the glory of last light, combines with the saturation and contrast of film to give us a uniquely unreal sense of these places. As film's distortions started to be stripped away for me, and with my long-term appreciation of pastel color, I am again getting comments that my work doesn't look like photography. With the digital component added to the picture, the first assumption often is that I've somehow manipulated the digital file into this very different look and feel. I try to be as straightforward as possible, letting whatever original vision I record of the place take precedence. I try to make a print as faithful to the original scene as possible. That's hard enough to do. The natural world is a stunningly beautiful, varied and subtle place. It doesn't need manipulation after the fact. It doesn't need embellishment. It is already self-embellished.
Koenderink: The Light Field and Shading in Pictures
"Shading" stands for diverse photometric effects, including shadowing, vignetting, the generation of texture for surface roughness and of course "shading" proper. All these have been effectively used by artists to induce pictorial relief and a sense of (pictorial) light field that pervades pictorial scenes. Classical shape from shading considers only shading proper, thus rendering its "problem" very hard. Human observers use illumination induced texture to perceive the flow of light over surfaces, thus trivializing the shape-from-shading problem. Recognizing this one can design more biologically valid shape-from-shading schemes. I will introduce such light transport issues and present psychophysics on the perception of the "light field" and the "flow of light" as revealed through texture.
Kubovy: On Mondrian, Balance, and the Perceptual Foundations of Art
John H. Brown, of the Department of Philosophy at the University of Maryland has thoughtfully analyzed the question of balance in Mondrian's art. He writes: "It is a commonplace about the classic phase of Mondrian's oeuvre that the artist achieves a subtle asymmetrical balance in his compositions. Much of our pleasure in engaging with these works seems to be appreciation of their balance, a balance which provokes our interest and awakens our admiration because it is so challenging. It strikes us not just as asymmetrical, which is not in itself remarkable, but as surprising, eccentric, improbable, even mysterious. The compositions are typically experienced as strong or stable yet also as dynamic, as containing the 'equilibrated contrasts' at which the artist aimed. In designs so purified of representational (pictorial) significance, so resolutely flat, so deprived even of the optical space found in Pollock or de Kooning, so limited in form, color and texture, what indeed would there be to admire were it not for some extraordinary virtue of balance? The artistic challenge in such a program comes in discovering in the compositions subtlety and depth commensurate with their presumed eminence." The purpose of this paper is to explore these interesting questions by first examining possible ways of answering them; then by reviewing the small literature that studies Arnheim's idea of the power of the center, the perceptual balance of colored areas, and the perception of Mondrian's work; thirdly by describing preliminary results of an experiment that might lead to a clarification of the notion of perceptual balance. Finally, I will discuss some foundational issues---I will distinguish between the study of psychological aesthetics and the study of the perceptual foundations of artistic perception and conclude that we are better equipped to pursue the latter. I may even hint that empirical psychological aesthetics is an inherently flawed project.
Lyon: Outside-the-box Camera Analysis
Even an ideal camera needs a finite aperture area to collection enough photons and to get enough diffraction-limited resolution, and therefore has finite depth of field, finite resolution, finite noise, and finite motion blur. Analyses of camera optical properties traditionally involve complicated equations (using f-number, focal length, and format size) that obscure some fundamental scaling relationships. But if the camera is viewed instead in terms of only its outside parameters, aperture diameter and field of view, some of these relationships become much easier to derive and simpler to understand. This method was used by Moritz von Rohr in the late 19th century, and was known to early television engineers, but has largely been lost from teachings of photographic optics, and deserves to be rehabilitated.
Mamassian: Perceived Shape from Shading and its Use in Photography and Painting
Chiaroscuro has been a powerful tool used by artists not only to depict the
three-dimensional shape of an object, but also to settle the mood of a scene.
Shape-from-shading has become a classical example of under-constrained problem
in computer vision, and while numerous creative solutions have been proposed,
few seem relevant to model human vision. Psychophysicists have studied the
ambiguities inherent to shading and shadows, and have revealed consistent biases
in the interpretation of these ambiguities. While shading is obviously important
in the appreciation of an image (a two-tone picture looks massively different
from the original grey-scale), it is still not clear to me what exactly the
human visual system extracts from the shading information. I will open the
discussion with an analysis of some shading anomalies that artists have
introduced in their paintings, voluntarily or not.
Palmer: Aesthetic Science: The Psychophysics of Spatial and Color Composition
Visual artists of all stripes continually face the problem of how to compose their works in aesthetically pleasing ways. Despite its importance and generality, the perceptual basis of aesthetic response has received inadequate empirical attention. I will report the results of a series of experiments that investigate aesthetic preferences for spatial and color composition. Strong, consistent preferences emerge in the spatial composition of simple images containing single objects or configurations of objects. Related experiments on judgments of "goodness of fit" for probe shapes (small circles or triangles) within a rectangular frame support similar effects, with striking evidence for the role of symmetry and balance in spatial composition. In the color domain, preferences for individual colors and color combinations are more ideosyncratic, but related to more basic colorimetric measurments. Contradicting artistic color theory, color harmony appears to be primarily based on perceived color similarity with no role for complementary colors. Still, some people prefer harmonious combinations more than others. The results show that "aesthetic science" is not an oxymoron, but a new and exciting topic within vision science that can shed new light on the nature of visual perception.
Perona: A Perspective on the Perfect Portrait
I will present evidence that artists, including those living during the Renaissance, combine multiple viewpoints in the same portrait. I will argue that this practice allows the artists to control the emotional valence of the portrait. I will present evidence from psychophysical experiments, as well as an analysis of the historical record.
Tyler: The Truth about Curved Perspective
In art circles, it is widely held that the rectilinear perspective is just an approximation to the true nature of perspective, which is curved in a spherical frame of reference. The rectilinear geometry of geometric perspective is perfectly accurate, even for a viewer with a curved retina. This can be seen by considering the center of projection for constructing the perspective as the eye’s optical center. Then the projection of lines in space to lines in the picture plane regenerates the exact geometry of the visual scene as it enters the eye, regardless of the shape of the retina. However, numerous factors may mitigate the purity of this geometric concept. If a perspective picture is viewed from a different distance (or direction) than the center of projection, the perspective appears distorted and conforms with some concepts of spherical geometry. If non-geometric aspects of depth cues are not consistent with the perspective construction, ambiguities may again result. Indeed, the fact that the painting is on a flat surface (as evidenced by convergence and accommodation) introduces a cue-conflict situation that can distort the depth impression. All these cues pertain to flat pictures, and do not constitute what Leonardo called ‘natural perspective’. As interpreted by Escher, this term refers to the combined impression of projection onto the curved retina as the eye views different parts of the visual scene. This is a deeper issue that cannot be addressed by images on a flat plane. In this concatenated view, parallel straight lines have two vanishing points viewable by moving the eyes, whereas on a flat plane they have only one. Paintings of curved perspective that attempt to capture these properties date back to Fouquet and Mantegna in the 15th century. They should be viewed as attempts to capture the true experience of natural perspective rather than as an improved geometry of perspective projection.