Does the Introduction of External Lighting Cues and Angle Affect Perception of Shape From Shading?

By Emily Huang

In Ramachandran's paper, Perception of Shape from Shading, he shows that the visual system places certain constraints on our perception of the world. Specifically, in cases where lighting cues are ambiguous, we seem to assume that light is coming from above, as it usually is in nature. He demonstrates this with shaded circles - when we view circles that have shading on the bottom, they appear to be convex, while circles with shading on top appear to be concave, which would only happen if we assumed light was coming from above.

This suggests that, with the addition of external lighting cues, the illusion should change - if there were external light coming from below, the circles with shading on top should appear convex, while the circles with shading on the bottom should appear concave. This is what I aimed to study in my illusion: how does the addition of external lighting cues affect our perception of shape from shading? And if the illusion stays the same, why? Additionally, I tried to see whether the angle at which one viewed the circles mattered. To imitate an external lighting cue, I overlaid a gradient in different directions on the original image of the shaded circles with a gray background. Afterwards, I adjusted the angle using an Instagram editing feature.

Below is the original illusion:


The illusion with an above external lighting cue:


The illusion with an below external lighting cue:


The illusion with a below external lighting cue and a downwards angle:


The illusion with a below external lighting cue and an upwards angle


Here is the link to the paper: Perception of shape from shading

Comments


Maddie C

It's always easier for me see the circles with white on top as the bumps, although I can flip it when the gradient is the right way. Comparing to my experiences with the other shape-from-shading illusions, makes me wonder about the importance of local proximity in these

Anonymous

@Maddie C,
@Katarina Bulovic,
@Jeremy Ma,

Yeah, for me it was the same - it was easier for me to see the circles with white on top as bumps, and with the angled photos, it appeared to be bistable - I could see the circles as both convex and concave. I think this answers my initial question of whether external lighting cues affects the perception of the original illusion; if the cue is coming from above, it deepens the illusion, and if it comes from below, the results are ambiguous, but still skewed towards light coming from above. The angle seems to also affect it a little, but not enough to completely reverse the illusion.

As far as changes go, I actually did try a real-world scene where light was shining from above or below, but it didn't do anything. It would be interesting to try to introduce conflicting cues, both digitally and in the real world. Also, I think it would be interesting to redesign the initial illusion - I believe the persistence of us perceiving the circles with shading on the bottom as convex may be due to the way the illusion is originally designed, since in real world scenarios, 3d bumps and craters probably wouldn't look like that.

Jeremy Ma

a)As a participant, the presence of external lightning cue amplified the effect of depth and tilting the images slightly altered the deepness of the circles. However, I think there is a bias to perceiving the more common shading as 'convex' when I looked at the original image, but with lightning it is more obvious.

b) As a researcher, I think that the stronger effect of depth that comes along with external lightning makes sense as it makes the prior of 'light from somewhere' more likely. One interesting effect might be having conflicting external lighting. Also the tilting of the plane and the reduce amount of depth also makes sense as the same gradient on a tilted plane would imply a shallower change.

Katarina Bulovic

a) As a participant, the external lighting cues actually don't seem to affect my perception of shape from shading. The second image did appear to be different, but I think that's because it's been flipped upside down (the light/dark parts of each circle have been flipped). The shape from shading of the angled pictures was slightly more ambiguous, but the same circles appear to be convex and concave.

b) As a researcher, I wonder if the reason my perception did not change is because the external lighting cues would need to be stronger. Maybe in a real-world scene where you could have a light actually shining from below the shape from shading perception would change.