How does an implied global 3D lighting affect perception of local shape from shading?

by Ben Radovitzky

As we saw in class, it is easy to detect 3D shape from shading by knowing a few simple things. First, surfaces tend to be approximately Lambertian, which means that the amount of light reflected towards our eyes depends on the angle of that light with the surface it is hitting. Second, we assume that light sources tend to come from above, because that is what we are used to seeing in our day to day lives. Together, these effects create the illusion we see in the circles/craters example from class. The circles with lighter shading at the top and darker at the bottom appear to be bumps, while the ones with lighter shading at the bottom and darker at the top appear to be craters. The question I ask is the following: what happens when other cues are given that imply some global illumination? For example, if I set a gradient of light to dark shading from bottom to top, does this affect our perception of whether the circles are bumps or craters? What if there is another local cue that happens to be a lot more salient (a face lit from the bottom is much more recognizeable than a simple circle without much detail)? Pay attention to whether or not the different circles either appear to pop out or sink into the page based on the background and cues.


Here we see the standard example we saw in class with circles that appear to bulge out of the screen if the top is lighter than the bottom, and appear to sink into the page when the bottom is lighter than the top.


In the left example, I tried to create the illusion of a "top" light source by putting a gradient on the background. I noticed that it makes the effect seem to pop a little more, but didn't change the nature of it. In the second image, I kept the same circle layout but inverted the gradient on the background to see if it would invert the percept of which circles were "popping out" versus "sinking in". I personally don't notice a huge difference unless I really try to focus on it.


In this final example, I used the exact same image as the first example but with a face in the bottom right that has a strong implication of some sort of "bottom lighting". I personally noticed that although this works fantastically at a local level, making the circle right above the face to appear popping out although the shading is "inverted", it works less well for the inverted circles that are farther from the face, which is interesting.

Inspiration: Ramachandran, V. S. (1988). Perception of shape from shading. Nature, 331(6152)

Comments


Maddie C

I find it easiest to see bumps for the. circles with white at the bottom when there's the gradient with white at the bottom. The face seems to influence me in seeing the circle immediately above as a bump. But strangely I have a hard time telling what is happening to the circle to the left of that... it's not clear to me that it's a. crater (it should be said in general I have a hard time seeing crater at all in these kinds of illusions)

Elian Malkin

a) For the image with a top-light, bottom-dark gradient, the shape from shading illusion persisted but did not seem any stronger than the original. The circles in the high contrast regions (either at the very top or bottom) produced a weaker perception of 3D shape. The image with a top-dark, bottom-light gradient strongly affected the percept for me, making the shape ambiguous. The circles with the top-down gradient appear somewhat spherical. The image with the face lit from the bottom did make the inverted circle above the face pop out, but did not affect the other circles.

b) The illusion does answer the question of whether global 3D lighting informs shape from shading. Lighting that is similar to natural light - top down - is effective at forming the perception of local shape from shading, while the opposite is true for unnatural bottom-up lighting. Subtle lighting cues such as the small image of a bottom-lit face can influence local shape perception but are not sufficient to undermine the top-down prior of natural light.

Ben Radovitzky

@Elian Malkin,
c) Your experience viewing my illusion answers my question well, as you state in part b). Global cues like a background gradient have the potential to slightly alter the strength of the effect, but are not quite enough to fully invert it.
d) I think that I would have liked to spend a little more time making the whole illusion seem more realistic. I feel like in this sort of case, where we are interpreting a "real" light source, it is useful for it to actually look like something that is being lit properly, so I think that putting a little more effort into making the lighting effects seem realistic would have made a large difference. I would have wanted to spend more time learning the GIMP software.

Nicholas Guiliano

a) In the first two experimental cases, I defintiely was able to see the bottom-up gradient pop-out "more" but the top-down gradient almost appeared to sink into the page less (especially the top-left most in the first gradient background example). In the example with the face lighting implication, I also experienced the local effect, but I also could extend it further out if I only looked at the circles with the same gradient type. Weirdly, if I looked at a different circle (the kind that should usually pop-out), it "reset" the information and I was unable to see the effect anymore.

b) These results indicate that the background implication of light source does impact perception of the pop-out, sink-in of the circles. While the effect might not be drastic for every case, the non-zero effect indicates that the light source plays a part in some mental analysis leading to the final percept.