Upload & Sell: Off
| p.1 #12 · In search of the 3D/MF look... |
I have been experimenting with this for a while using a Gigapan Epic Pro robotic panoramic head. I've done the following observations:
1) Too close is no good. It's not just about shallow DOF. You can get it easily enough with a macro lens on a small sensor. To make things interesting and to get a 'large format' look you need to have some distance to the subject.
2) At medium close distances you get the desired effect in terms of foreground/background relation and a unique look but field curvature and parallax is a big problem.
3) Using the lenses wide open isn't always a good solution due to vignetting. Even with lens correction profiles in lightroom it doesn't
Here are three examples of the same scene but with different lenses.
A) 100 mm (Zeiss 100 MP @ f/2), 18 images:
B)135 mm (Canon 135L @ f/2), 40 images
C) 200mm (Canon 70-200/II @ 200mm, f/2.8), 84 images:
Now these types of shots are more experimentally interesting than actually useful. You can see the extreme field curvature and the problems with the projection.
However, the technique becomes very interesting at larger distances where both field curvature and parallax are small. The look isn't as dramatic as in the images above, but the effect is more subtle, and really much more in line with a 'large format' look.
70-200 @ 200mm, f/2.8, ~200 images but cropped
100 MP, f/2, ~30 images, also somewhat cropped
There are two other benefits of using a stitched panorama and both come from the fact that you end up with huge images. When you resize the panorama to 'normal' size, i.e to the equivalent of what you would get from a single frame from the camera you get much better noise performance and better color accuracy. In a single frame each pixel only has one real color value - the rest are interpolated from nearby pixels. When you resize a larger image you'll get much better averaging per pixel and subsequently higher color accuracy. At the same time you can use very high ISO with no visible noise in the resized image.