Stacking, stitching, and the entrance pupil

A forum to ask questions, post setups, and generally discuss anything having to do with photomacrography and photomicroscopy.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

elf
Posts: 1416
Joined: Sun Nov 18, 2007 12:10 pm

Stacking, stitching, and the entrance pupil

Post by elf »

When stacking and stitching, how do you determine where the entrance pupil is located? Does it matter when you're doing an orthographic stitch (moving the camera or subject laterally)? All of the stacked macro panos I've done have moved the focus point by focusing the lens instead of moving the camera. The entrance point when the lens is focused at mid-depth seems to work fairly well, but there's always a little parallax error.

When stacking and moving the camera, how do you determine how much each individual move should be?

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Ed,

The entrance pupil is always located at the apparent position of the limiting aperture, as seen by the subject. This is the same for macro work as for traditional panorama stitching.

But I think the question you're really asking is, how does one minimize parallax errors while stacking and stitching?

The best strategy is always the one that causes the entrance pupil to move the least angle, as seen by any point on the subject.

Ideally, the entrance pupil will not move at all, in which case there will be no parallax errors.

This is achieved by telecentric optics by effectively pushing the entrance pupil to infinity. This allows the camera+lens system to be scanned orthographically across the subject while the entrance pupil sits in the same position, out there at infinity. (It's bizarre to watch, if you look into the front of the lens system as the subject would.)

Zero movement of the entrance pupil can also be achieved with lens on bellows, by locking the lens mount in position with the entrance pupil at the center of rotation, and focusing by moving the camera back and forth on the rails. In theory you can do the same thing without the bellows, but coordinating movement of a macro rail and a lens focusing helix becomes a nightmare.

With most other setups, the entrance pupil will move as you refocus and scan, and the challenge is how to minimize the movement.

Assuming that you focus by moving the lens+camera as a unit, then using a longer lens will help because that generally puts the entrance pupil farther away from the subject. The total amount of z-axis movement will be the same regardless of lens focal length, so placing the entrance pupil farther away results in proportionally less angular movement as you refocus. (One way to think about telecentric systems is that they just carry this principle to its limit of zero angle when the entrance pupil is infinitely far away.)

At high magnification (>>1:1), the only practical approach is to focus by moving the camera+lens as a unit, since changing the lens-to-camera distance has only a small effect on focal distance.

At low magnification (<<1:1), the only practical approach is to focus by changing the camera-to-lens distance. (Imagine the other way, trying to move focus from 1 meter to 2 meters by moving the camera+lens as a unit.)

At moderate magnification, say 1/4X to 1X, both approaches are practical. Which is better depends on the optics, and I think is best determined by experiment.

In any case, the best rotation point is probably at the center of the range of movement of the entrance pupil. Exception: at sufficiently high magnification, the DOF per frame becomes so shallow that you can pretend the images have orthographic perspective, even though they really don't!

The amount of refocusing per step is again best determined by experiment. What you want is to be sure that every subject feature is well focused in at least one frame. If every subject feature is well focused in two or more frames, then you have stepped too closely. With proper image registration in the software (which is now done by all the packages I know of), there is no penalty in image quality from stepping too closely. It just takes longer to shoot and to process. Generally speaking, the required focus step at optimum aperture varies inversely as magnification squared, so doubling your magnification (2X) shrinks your focus step by 4X. But calculating it accurately is a pointless exercise. In this digital age, it's much simpler and more reliable to just shoot a test series with a very small step, and look through it to see what larger step meets the criterion of every feature being well focused in at least one frame.

Did this answer your questions? If not, please ask again.

--Rik

elf
Posts: 1416
Joined: Sun Nov 18, 2007 12:10 pm

Post by elf »

I was curious how everyone was determining where the entrance pupil was located when shooting between 1:1 and 1:4? It would be nice to be able to take two images, measure the parallax error, then calulate where the entrance pupil is located relative to the current setting.

Olympus claims their digital lens are telecentric, but the FOV changes dramatically when changing the focus point, so I don't think they're using the same definition of telecentric :)

I need to try the camera movement technique to see how it compares to refocusing with the lens.

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

elf wrote:I was curious how everyone was determining where the entrance pupil was located when shooting between 1:1 and 1:4? It would be nice to be able to take two images, measure the parallax error, then calulate where the entrance pupil is located relative to the current setting.
I don't know how (or even whether) anybody else does it. On those rare occasions when I want a fairly accurate measurement, I point a second camera into the front of the first (shooting) camera. Stop down the first camera. Focus the second camera on the aperture of the first. You're now focused on the plane of the entrance pupil. Slide the first camera sideways out of the way, position a marker so that it's in focus, and use that to mark a piece of masking tape on the first camera's lens housing. It's not perfect, but it'll be on to within DOF of the second camera, which is often a couple of mm.
Olympus claims their digital lens are telecentric, but the FOV changes dramatically when changing the focus point, so I don't think they're using the same definition of telecentric.
They almost certainly mean "telecentric on the sensor (image) side", as opposed to "telecentric on the object (subject) side" like I'm talking about. Being telecentric on the sensor side is a good thing with digital sensors. It means that the ray cones hit the pixels in the middle of the sensor in the same way that they hit the pixels in the corners. It's the same concept -- ray cones perpendicular to an image plane -- but applied to a different image plane (sensor versus object).

Lenses that are telecentric on the object side are not much use for normal photography, since they look straight forward and can only focus on objects that are smaller than the lens diameter.

--Rik

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Just filling in some links here, to make this topic a better reference for questions about how best to shoot frames for stacking.

Problem (or not?) taking frames for stacking with CombineZM

Telecentric optics, third round

Stack-and-stitch: ordinary optics vs telecentric

--Rik

Post Reply Previous topicNext topic