Vancouveria planipetala - Inside-out Flower

Images taken in a controlled environment or with a posed subject. All subject types.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

PaulFurman
Posts: 595
Joined: Sat Oct 24, 2009 3:14 pm
Location: SF, CA, USA
Contact:

Vancouveria planipetala - Inside-out Flower

Post by PaulFurman »

This is a very small flower, about 5mm wide so this was about 3 or 4x on full frame. Canon 35mm MP-B at f/4, stack of 100 frames. Full crop below and the first links to full original.

Image

Image

PaulFurman
Posts: 595
Joined: Sat Oct 24, 2009 3:14 pm
Location: SF, CA, USA
Contact:

Post by PaulFurman »

I played more with the telecentric idea on a 50mm f/1.4. It has such a large rear element, I put black paper with a hole on the front and reversed it. First try: the stop was about f/1.8 and that looked bad. The ones below had a hole that looked like f/2 from the subject. Then I learned f/2.8 on the aperture ring looks better*, the f/2 at lower magnification has problems with foreground obstruction haze, I'm not sure why other than a deep subject and wide aperture. I haven't tried with a smaller added aperture.

Theory says a reversed normal lens should perform better with longer bellows extension and it does in my judgment work better at 4.5x max bellows than 1.7x min bellows shown.

*Note that stopping down the aperture ring further removes the telecentric effect and the subject once again changes scale while moving the focus rail requiring cropping or heavy scaling


50mm f/1.4 Ai at f/2.8, 1/2 sec. reversed with front telecentric aperture added equal to f/2 from front, stack of 170:

Image


50mm f/1.4 Ai at f/2 1 sec. reversed with front telecentric aperture added equal to f/2 from front, stack of 48:

Image



Then 2 shots with the Olympus 20mm f/2, 1/6 sec. and 1/2 sec. min bellows 6x. It did pretty well with the bellows closed but did not hold up well at higher magnifications:

Image

The shot above is the back of the flower. The stem (with glandular hairs) almost completely dissolves in the depth of field where I stopped the stack of 106. Below is a stack of 48.

Image

ChrisR
Site Admin
Posts: 8662
Joined: Sat Mar 14, 2009 3:58 am
Location: Near London, UK

Post by ChrisR »

Note to self = look better at tiny flowers! Some pics ( crummy by comparison) here
Do all the stacking artifacts retouch out OK - particularly where the stem(?) comes in behind the petal(?) on the right?
If so it would make a nice print, a metre or so wide might do it justice? :)

Edit - hadn't seen beyond the first image. Certainly looks the sharpest.
Scaling problems? I suppose a longer lens would help?

PaulFurman
Posts: 595
Joined: Sat Oct 24, 2009 3:14 pm
Location: SF, CA, USA
Contact:

Post by PaulFurman »

ChrisR wrote:Note to self = look better at tiny flowers! Some pics ( crummy by comparison) here
Hey, those are my pics :P first hit on google :)
Do all the stacking artifacts retouch out OK - particularly where the stem(?) comes in behind the petal(?) on the right?
The stem behind in the first shot turns blurry close to the petal. There's nothing to recover. The low mag shot has these blurry backgrounds near the edges worse.
Do you need telecentricity for the later pics?


The front view was cropped because the petals below came too close, which changed magnification and streaked as I stepped back. The back view has a thin overall depth so it was fine. I don't know if the telecentric helps or hinders yet because there is a price to pay in overall sharpness but this one is not bad.

ChrisR
Site Admin
Posts: 8662
Joined: Sat Mar 14, 2009 3:58 am
Location: Near London, UK

Post by ChrisR »

Oops sorry didn't mean to call your images crummy, just nowhere near as wonderfully revealing!!
There's nothing to recover.

Similar with stem on right of first pic of second post, which looks as though it must be in the middle of the stack somewhere?? I've often got these and wondered if I was doing something wrong.
I don't know if it's an inevitable effect from stacking, or if some stacking programs would have a variable which could be used to minimise it. ??

(er, Rik, :? )

rjlittlefield
Site Admin
Posts: 23543
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

(er, Rik, :? )
Given a good algorithm, loss of background detail next to a foreground edge is due to occlusion. When the background is (or would be!) focused, it is partially or completely obscured by the foreground element.

The exact effect that you get depends on details of how things line up when the background is in focus.

Think about the cone of light that is accepted by the aperture.

If half of that cone is obscured by the foreground, then the light rays that get into the lens come half from the in-focus background that you want to look at, and half from the out-of-focus foreground that you don't. In that case, background detail is still available (and will be found by PMax), but the contrast is diminished. If the foreground is bright, and the background is dark, then the contrast will be diminished by much more than half because you're adding a lot of OOF light to just a little light from focused details.

As more of the cone is obscured by the foreground, the problem gets worse. If the entire cone is obscured by foreground at the time the background would be in-focus, then there simply is no detail available. In this case the pixels where you want to see in-focus background are inevitably filled with OOF foreground.

Of course "how things line up" is a matter of perspective as well as subject geometry. If the perspective never changes, so that things line up the same way regardless of where you're focused, then in most cases the worst that happens is that half the aperture gets occluded. But if the perspective does change, then the occlusion can get arbitrarily bad.

With most lenses, perspective changes if you move the lens. How much it changes depends on how deep the subject is, compared to where the entrance pupil is. If the entrance pupil is close and the subject is deep, it's easy to get complete occlusion of background near foreground edges.

There are several ways to minimize or avoid this. One of them is elf's approach -- leave the front lens in one place and focus by changing the bellows rear extension. This provides exactly zero change in perspective, no matter what lens you use. A second approach is to use a longer lens, or a telephoto lens that has the entrance pupil pushed back. This reduces the change in perspective, but does not eliminate it. A third approach is to use a telecentric system, either by adding just an aperture as Paul did, or by assembling a telecentric combo. If the system is perfectly telecentric, then this provides exactly zero change in perspective. If it is not perfectly telecentric, then the change will be non-zero but much reduced compared to an ordinary lens.

Each of these approaches has advantages and disadvantages. Focusing by rear extension uses the lens essentially as it was designed, which minimizes aberrations. But it produces relatively large changes in magnification, which requires extensive cropping and as a result may reduce your effective resolution. At moderate magnifications, it may also be difficult to make bellows movements precise enough to place the focus steps where you need them. Using a longer lens or a telephoto design also minimizes aberrations, but does not completely eliminate perspective changes. Using a telecentric system solves the perspective problem, but generally maximizes aberrations by selectively using outer portions of the glass for off-axis portions of the image.

Generally speaking, a telecentric combo will work better than just adding an aperture. The best telecentric combo is to place a weak achromat close to the subject, working in combination with a long lens and a correspondingly long extension farther back. (See the setup HERE.) In this case the nearby achromat just serves to align the ray cones to provide telecentricity, while the main lens farther back does the heavy lifting to form a well focused image.
ChrisR wrote:
There's nothing to recover.

Similar with stem on right of first pic of second post, which looks as though it must be in the middle of the stack somewhere??
That stem is a great example because it illustrates two different effects.

At the bottom right end of the stem, where it passes behind the horizontal edge of the petal, there actually is detail right up to the edge, but with degraded contrast. This is a case where the cone was only partly occluded.

At the upper left end of the stem, where it passes behind the vertical edge, there appears to be no detail at all. This is probably a case where the cone was completely occluded.

If the optics were perfectly telecentric, then the maximum occlusion would be only half the cone, and the upper left end of the stem would look the same as the lower right -- detail present, contrast degraded.

The fact that there appears to be complete occlusion behind the vertical (tangential) edge suggests that the optics were not completely telecentric. I see some radial dust trails that indicate scale changes.

But it's also possible that the software was being misled by focus blobs into making scale adjustments that really were not needed and were actively counterproductive. When using telecentric optics, it's best to turn off scale adjustment in the software
ChrisR wrote:I've often got these and wondered if I was doing something wrong.
Partial occlusion cannot be avoided. Complete occlusion can be avoided by proper choice of lens and focusing technique, but sometimes it's a lot of trouble.
ChrisR wrote:I don't know if it's an inevitable effect from stacking, or if some stacking programs would have a variable which could be used to minimise it. ??
In theory, it's possible to computationally remove the effect of partial occlusion. But it's a very difficult problem because to accurately estimate the contamination, you need to know both subject geometry and optical geometry -- essentially, where the cones are and what is seen by them. This would be bad enough if lenses were perfect and apertures were round, but they're not. This is back to bokeh -- what does an OOF point source look like? I'm not aware of anybody who has tackled this problem even as a research effort.

--Rik

lauriek
Posts: 2402
Joined: Sun Nov 25, 2007 6:57 am
Location: South East UK
Contact:

Post by lauriek »

I think we've discussed this before Rik! ;)

I was just wondering after reading and re-reading your last reply, okay this problem is /far/ too large to come up with a generic solution. But, would it be possible to do some studies on say a couple of our favourite lenses (e.g. the Oly 38/2.8 and the Nikon 10x/0.30 CF) - and come up with some functionality which would help?

I suspect not as you said "to accurately estimate the contamination, you need to know both subject geometry and optical geometry" - so you'd have to understand the subject geometry as well, though the 3d thing you do with Zerene suggests you /might/ be able to work out something about the subject geometry (maybe with a little help from the user) which could help with this?

Sound like a project? :D

PaulFurman
Posts: 595
Joined: Sat Oct 24, 2009 3:14 pm
Location: SF, CA, USA
Contact:

Post by PaulFurman »

You deduced correctly that my setup is not completely telecentric. The view from the subject places the the entrance pupil only about a foot behind the lens but it looks like almost no scaling is occurring through the viewfinder. If I stop down, this isn't the case and things scale dramatically. This is such an easy modification to make and works reasonably well. The aperture goes right over the front element, then the BR2 reversing adapter holds it there.

It seems to me this telecentric thing makes the occlusions worse. I'm not at all sure but it could be that without it, the foreground objects will grow by the time you get to them and obscure the worst areas. That's likely wrong though :-( However this is the worst case I've seen of this phenomenon.

The simple lesson, particularly for my low mag shot with the wost effects, is... I could have changed my position to compose it to minimize overlap. Pretty simple in this case anyways.

PaulFurman
Posts: 595
Joined: Sat Oct 24, 2009 3:14 pm
Location: SF, CA, USA
Contact:

Post by PaulFurman »

rjlittlefield wrote:
ChrisR wrote:
There's nothing to recover.

Similar with stem on right of first pic of second post, which looks as though it must be in the middle of the stack somewhere??
That stem is a great example because it illustrates two different effects.

At the bottom right end of the stem, where it passes behind the horizontal edge of the petal, there actually is detail right up to the edge, but with degraded contrast. This is a case where the cone was only partly occluded.

At the upper left end of the stem, where it passes behind the vertical edge, there appears to be no detail at all. This is probably a case where the cone was completely occluded.

If the optics were perfectly telecentric, then the maximum occlusion would be only half the cone, and the upper left end of the stem would look the same as the lower right -- detail present, contrast degraded.

The fact that there appears to be complete occlusion behind the vertical (tangential) edge suggests that the optics were not completely telecentric. I see some radial dust trails that indicate scale changes.
The right side, I started the stack there so the fade off doesn't go any further. Here's a very short video showing that stack. It's possible I had a bit of 'reverse perspective' working against me here.

ChrisR
Site Admin
Posts: 8662
Joined: Sat Mar 14, 2009 3:58 am
Location: Near London, UK

Post by ChrisR »

An obvious possibility is to just light the plane in focus, must have been done. New query here:
http://www.photomacrography.net/forum/v ... 4698#54698

Post Reply Previous topicNext topic