Hack23.See How Brightness Differs from Luminance: The Checker Shadow Illusion
|
Hack 23. See How Brightness Differs from Luminance: The Checker Shadow Illusion
A powerful illusion of brightness shows how our brain takes scene structure and implied lighting into account when calculating the shade of things. A major challenge for our vision is the reconstruction of a three-dimensional visual world from a two-dimensional retinal picture. The projection from three to two dimensions irrevocably loses information, which somehow needs to be reconstructed by the vision centers in our brain. True, we have two eyes, which helps a bit in the horizontal plane, but the vivid self-experience of seeing a 3D world clearly persists after covering one eye [Hack #22] . In the process of reconstructing 3D from 2D, our brain cleverly relies on previous experience and assumptions on the physics of the real world. Since information is thus fabricated, the process is prone to error, especially in appropriately manipulated pictures, which gives rise to various large classes of optical illusions. We will concentrate here on a fairly recent example, Ted Adelson's checker shadow illusion.1 2.12.1. In ActionTake a look at Adelson's checker shadow illusion in Figure 2-19. Figure 2-19. Adelson's checker shadowwhich is brighter, A or B? We would all agree that one sees a checkerboard with a pillar standing in one corner. Illumination obviously comes from the top-right corner, as the shadow on the checkerboard tells us immediately (and we know how important shadows are for informing what we see [Hack #20] ). All of this is perceived at one rapid glance, much faster than this sentence can be read (lest written!). Now let's ask the following question: which square is brighter, A or B? The obvious answer is B, and I agree. But now change the context by looking at Figure 2-20. The unmasked grays are from the two squares A and B, and unquestioningly the two shades of gray are identical (in fact, the entire figure was constructed just so). Figure 2-20. This checkerboard is the same as the first, except for the added barsnow does A look brighter than B? You can prove it to yourself by cutting out a mask with two checker square-size holes in it, one for A and one for B, and putting it over the original checkerboard (Figure 2-19). 2.12.2. How It WorksIf squares A and B in the first case have clearly differing brightness and in the second case they have the same, what gives? Surely the two alternatives exclude each other? The solution in a nutshell: brightness depends on context. There is a good reason that visual scientists describe their experiments using the term luminance rather than brightness. Luminance is a physical measure, effectively counting the number of light quanta coming from a surface, then weighting them by wavelength with regard to their visibility. (The unit of measurement, by the way, is candela per square meter, cd/m2. A candela was originally defined as the light from a standard candle 1 foot away.) Brightness, on the other hand, is a subjective measuresomething your brain constructs for your conscious experience. It depends on previous history (light adaptation), the immediate surroundings (contrast effects), and context (as here). It has no dimension but can be measured using psychophysical techniques.
What exactly is happening when comparing Figure 2-19 and Figure 2-20? Well, when I initially asked, "Which square is brighter?", I knew you would give the deeper answer, namely the lightness quality of the substance the squares are made of. I knew youor your smart visual systemwould assess the scene, interpret it as a 3D scene, guess the shadowed and lit parts, predict an invisible light source, measure incoming light from the squares, subtract the estimated effect of light versus shadow, and give a good guess at the true lightnessthe lightness that we would expect the checker squares to really have given the way they appear in the scene they're in. With the mask applied (Figure 2-20), however, we create a very different context in which a 3D interpretation does not apply. Now the two squares are not assumed to be lit differently, no correction for light and shadow needs to be applied, and the brightness becomes equal. The luminance of squares A and B is always identical, but due to different context, the perceived brightness changes. By the way: there are more places in that figure where luminances are equal, but brightness differs, and hunting for those is left as an exercise for the gentle reader. This striking checker shadow illusion by Ted Adelson teaches us quite a number of things: it demonstrates how much unconscious scene computation goes on in our visual brain when it applies inverse perspective and inverse lighting models. It shows us how strongly luminance and brightness can differ, giving rise to perceptual constancies, here light constancy. It also demonstrates the "unfairness" of the term "optical illusion": the first answer you gave was not wrong at all; in fact, it was the answer one would be interested in, most of the time. Imagine the checkerboard were like a puzzle, with missing pieces, and you had to hunt for a matching piece. Material property is what we need then, independent of lighting. In fact, estimating the "true" material properties independent of context is a very hard computational problem and one that hasn't been solved to a satisfying degree by computer vision systems. 2.12.3. In Real LifeCorrection of surface perception for light and shadow conditions is such a basic mechanism of our perceptionand one that normally operates nearly perfectlythat very artificial situations must be created by the accompanying figures for it to reveal itself. That is why we need technical help taking photographs: since photos are normally viewed under different lighting conditions compared to the original scene, professional photographers need to go a long way arranging lighting conditions so that the impression at viewing is the one that is desired. 2.12.4. End Note
2.12.5. See Also
Michael Bach |
|
EAN: N/A
Pages: 159