Joined: 01 Aug 2006
Location: Richland, Washington State, USA
|Posted: Sun Jan 13, 2008 12:46 am Post subject: Stereo pairs: the effect of lighting shifts
|From time to time I mention that to get the best stereo, you have to be sure that the lighting of the subject is the same in the left and right views. I thought it might be helpful to show an illustration.
What we have in this post is a fly face illuminated with a mix of lighting: part very diffuse and part directional. The diffuse component illuminates all portions of the face and is essentially shadow-free. The directional component illuminates primarily the left portion of the face and casts fairly sharp shadows.
I shot the face in both ways that are commonly discussed:
1. fixing the lights with respect to the camera, while the subject rotates to form left and right views, and
2. fixing the lights with respect to the subject, while the camera shifts to form left and right views.
What happens in case 1 (rotating only the subject) is that shadows move on the surface of the subject. Disparity in the left/right views is partly due to 3D structure, and partly due to mismatching shadows. The mismatching shadows amount to misleading information, which the viewer must either ignore or be confused by.
In case 2, shadows remain fixed on the surface of the subject, so disparity in the pair is due entirely to 3D structure. With diffuse or finely textured surfaces, most of the detail information ends up matching exactly. With very shiny structures, specular highlights will shift between the two views. However, this shift is the same type that we're used to seeing with shiny objects in our daily lives, so it's less likely to be misinterpreted.
I'm very fond of stereo, and for me the difference between cases 1 and 2 is quite obvious.
When the lighting is fixed to the camera (case 1), areas with mismatching shadows tend to "shimmer" and I have an unpleasant feeling of constant tension.
But when the lighting is fixed to the subject (case 2), the images fuse cleanly and effortlessly.
It's likely that other people are less sensitive to this problem. And of course, if you can't see stereo at all you're going to have trouble seeing much difference.
Anyway, here are the images, crossed-eye, first large, and then the same thing at half-size. Particular regions to notice are the left cheek (left to us; it's the fly's right), the bristles above the cheek and at left bottom, and especially, below-left of center in the bowl behind the "nose" formed by the antennae.
Case 1 (not so good)
Case 2 (much better)
Both together, at half size:
Shot with: Canon 300D with Olympus 38mm bellows macro lens, f/8, stacked at 0.010". The area shown is about 40% of frame width, 2.2 mm on the subject. "Directional" lighting means a 5 mm diameter fiber head, 120 mm away from the subject.