Stacking exercise

[chauncey]

A orchid stack, 12 slices done with PS CC/Helicon/Zerene (2 ways).

For some reason all three techniques boost the RGB levels 40-55% in the darker areas over the base
levels in PS CC. With my interest in flowers, muddy areas are my bane and they all show that problem,
sometimes in different places.

These 100% crop examples have had no repair work, were stacked with default settings, and are
representative of the whole image.

PS CC



Helicon



Zerene DMap



Zerene PMax



The grittiness in the PMax image puts me off but it does have the crispy edges, albeit noisy.
Regarding the brightness boost, it would seem that slight adjustment in exposure might be of benefit.

Which is better for my flowers, IMHO, it's a close call, as they all need repair...

[DanK]

Glenn,

A very interesting post. Thanks.

I had not thought about why, but my experience with both of my macro lenses is that changing focus at close distances does change magnification, so it must be that the nodal point moves.

A few thoughts about longer focal lengths:

--it matters whether you thinking about framing or magnification. If you are thinking about framing or reach, then you do have a longer working distance with a crop sensor camera. If you are concerned about magnification, you don't. You get 1:1 magnification at the same distance, regardless of the sensor size.

--I see your point about the lesser parallax difference with a longer focal length, but I wonder how much it matters in practice. I switched from a 50D to a 5D3 for most of my flower macros a year ago, and I haven't noticed any difference in stacking using Zerene, but then again, I haven't done any A/B tests. If you follow postings here, you will find a lot of excellent stacked images taken with short FL lenses, including MP-E 65s on FF cameras.

[Glenn NK]

Dan:

Yes, I've seen some excellent stacks done with the MPE-65, but I'm wondering if that lens has something else special about it - maybe a nodal point that doesn't move with focus distance.

Just wondering out loud.

Glenn

[rudi]

Although very interesting I can not ad very much to this technical discussion.

Keep in mind that I just started with the mpe and that all shots have been done semi-handheld. But what I can see so far is that the amount of the object seen also varies with the camera to subject distance.


I

[DanK]

Dan:

Yes, I've seen some excellent stacks done with the MPE-65, but I'm wondering if that lens has something else special about it - maybe a nodal point that doesn't move with focus distance.

Just wondering out loud.

Glenn

I'm just guessing too, but I would guess not. AFAIK, MP-E 65 stacks have to be done by moving the lens for changes in focus, so there has to be a parallax change. There are also a lot of 100mm stacks there too, but I don't recall how many are done with crops.

I'm way to busy to do it now, but at some point, I will try to remember to do an A/B comparison using my crop and FF, using the same lens.

[Glenn NK]

Although very interesting I can not ad very much to this technical discussion.

Keep in mind that I just started with the mpe and that all shots have been done semi-handheld. But what I can see so far is that the amount of the object seen also varies with the camera to subject distance.


I

Rudi:

I have no person knowledge of this lens (I've seen a few images created by it) but don't know how it focuses.

But if is used by moving the lens (as Dan notes), then it should be governed by the principles of the sketch I drew up.

This is the same principal that makes faces in selfies look fat and round, and why conventional wisdom suggested that a portrait lens for head/shoulders shots should be in the 80 mm range on a 24 x 26 mm film/sensor format.

The principles of optics/physics don't change much with time.

Glenn

[rjlittlefield]

Hello! I am Rik Littlefield, the fellow who wrote Zerene Stacker.

Rudi contacted me via support@zerenesystems.com to ask the same question that he started this thread with. He thought that my reply might be helpful to other readers of this thread.

Here is what I wrote in reply to Rudi's question:

Rudi,

This is a complicated situation. Probably it is explained in bits and pieces other places, but for this problem I think it is easiest if I write a new explanation here.

First, the short answer is that I'm not sure what all is going wrong. I would need to work with the original source images to be sure.

However, from the stacked results that I see here, I can make some educated guesses.

The first big issue is that Zerene Stacker and Photoshop use very different methods for aligning images.

Zerene Stacker uses only shift/rotate/scale, and its measure of good fit is to minimize the average difference in pixel values across the entire image. This essentially assumes that the images were shot from the same viewpoint and with almost the same framing, with the images varying only in focus.

Photoshop also uses shift/rotate/scale, and in addition it will introduce various non-linear transformations like barrel/pincushion distortion and perspective keystoning. Photoshop's measure of good fit is to minimize the average distance between "features" that are selected for having distinctive patterns of pixel values in a small neighborhood. This method has its roots in side-by-side panorama stitching. It essentially assumes that the images were shot from the same viewpoint but possibly have very different framing and only a little overlap.

When the assumptions are violated, both methods will fail, but they fail in different ways. In the current stack, it looks to me like either the camera or the leaves were moving around so that the leaves and the fly and the background line up differently in different frames. Photoshop appears to have latched onto the fly that you care about, while "letting go" of the leaves, especially the leaves in the upper left corner. Zerene Stacker, in contrast, has essentially chosen to keep the leaves lined up as well as it could, while letting go of the fly. (This is because the fly occupies relatively few pixel positions compared to the leaves.)

The second big issue is that Zerene Stacker and Photoshop have very different methods for combining images.

In Zerene Stacker, the PMax method is relentless about preserving the sharpest available detail at each pixel position. This works very well in a well formed focus stack where everything is properly aligned, but when there are misalignments, it results in severe "ghosting" as each sharp detail may be seen simultaneously in all the positions it occupies across all frames.

Photoshop, on the other hand, is committed to the concept of carving out fairly large regions of pixels, each from a single source image, and then pasting those regions together with a bit of blending. When there are misalignments, this method avoids ghosting, but instead introduces "steps" on the region boundaries.

These differences are evident in the two results that you sent to me. The ZS result shows ghosting; the CS6 result shows steps.

(As further information, Zerene Stacker has a second method, DMap, that is more like what Photoshop does. The major difference between DMap and Photoshop is that DMap assumes that the images are shot in depth order, either back to front or front to back. It uses this assumed ordering to figure out what to do in areas where there is no focused detail.)

The third big issue is that Zerene Stacker and Photoshop have very different methods for determining the final framing.

Zerene Stacker determines final framing by selecting a single source image, either the first or the last one in the input sequence, then using that one image's framing for the final result. The default method is to choose whichever end of the stack has the narrowest field of view, which tends to avoid an artifact known as "edge streaks".

Photoshop determines final framing by registering all the images against each other, then creating a large frame that encompasses all of the registered source images. Essentially this results in producing the widest possible field of view, potentially including lots of edge areas that are covered by less than all the source images.

These differences in framing are easily seen in the two results that you sent. It appears that there was substantial misalignment between the various frames, so that Photoshop produced a "wide angle" output covering all frames, while Zerene Stacker produced a "narrow angle" output determined by just one frame.

Finally, in Zerene Stacker there is an image-saving option labeled "Retain full dynamic range" that is often misunderstood. What this option does is to reduce contrast and brighten or darken the image as necessary to avoid clipping pixel values that have been internally computed to be either "brighter-than-white" or "darker-than-black". Your Zerene Stacker output has a generally low contrast appearance that is typical of this option having been selected when the output image was saved.

With luck, this discussion will explain why the two output images have the appearances that they do.

But I suspect that you're really interested in knowing how you can get better results in the future.

The answer to that question is to avoid movement. Movement of either the subject or the camera will make it impossible for any stacking software to produce a perfect result.

If movement cannot be avoided, then try to select your viewpoint so that everything visible in the frame moves together. (In the current case, I suspect that the leaves at upper left and lower right are moving differently.)

If there is significant residual movement, then in Zerene Stacker it's usually better to use DMap than PMax. That will avoid the ghosting problem. Even so, you may end up getting a better result from Photoshop in cases like your current stack, where Photoshop happens to latch onto the subject you care about and keeps it aligned at the expense of other image elements.

Note that "avoid movement" may mean avoiding some situations altogether. There are not many successful focus stacks of flies on leaves in complex environment. The reason for this is that almost always something moves between frames. This not only messes up the automatically stacked result, but usually makes it difficult to fix with manual retouching as well. In the end (in these cases) the result is not worth the cost. Photographers who do successfully stack flies on leaves in complex environments generally do that by getting up early and shooting while both the air and the bugs are quiet. There's a definite reason why so many stacked flies are covered by dewdrops or frost!

Getting back to your original question:
> What am I doing wrong with the stack in Zerene, only stacked in Pmax.

I think the problem is not that you're doing something wrong in processing the stack, but rather that it's an impossible stack in the first place because things are moving around too much from one picture to another.

Again, I would have to see the original frames to be sure, but this is my best guess given what I see in your outputs.

Did this discussion answer your questions? Please let me know. Thanks!

Best regards,
Rik Littlefield
Zerene Systems

I may be able to shed some light on a couple of other issues that have been raised in this thread.

1. Regarding magnification change and change in viewpoint (center of perspective).

As noted by DanK and Glenn NK, magnification change by itself is not a problem. Effective magnification changes almost all the time, no matter how you adjust focus, so all the stacking packages handle that as a matter of course. What does cause a problem is change in viewpoint (center of perspective), so that various parts of the scene shift positions with respect to each other.

This issue is surprisingly complicated, and bears on the issue of how one should adjust focus so as to get the best result. See http://zerenesystems.com/cms/stacker...ringversusrail for some discussion and illustration in the context of Zerene Stacker.

Glenn NK writes that

If the lens is re-focused for each shot, and the nodal point does not move (an impossibility perhaps?)

This is quite possible. All that's necessary is for the lens aperture and all elements in front of it to stay in one place. Some lenses do this naturally, because they focus by moving only elements in back of the aperture. Any bellows system that is focused by changing the rear extension works. There are also special lens designs called "telecentric on the object side", whose center of perspective is at infinity even though the lens is physically short and has completely ordinary conjugate distances for focusing. Telecentric designs are particularly valuable for stack-and-stitch at macro magnifications, because they allow the lens and camera to be shifted sideways without introducing parallax error.

As for the MP-E 65, there's nothing special about that lens except the fact that it operates at relatively high magnification. It turns out that the ill effects of moving the lens depend on the ratio of DOF to distance, and this ratio drops sharply with increasing magnification. By the time you get around to using microscope objectives, there's often no way to adjust focus other than by changing the subject-to-lens distance, but at the same time it doesn't matter because the DOF is so shallow. (Above 10X, I normally recommend turning off Scale adjustment because it's more likely to degrade the result than improve it.)

2. Changes in brightness, contrast, color, and noise.

In Zerene Stacker, all such changes can be prevented by
. turning off Brightness adjustment,
. using the DMap method, and
. not setting the option to "Retain full dynamic range" (which really means to apply a curves adjustment so as to preserve any "brighter than white" or "darker than black" values that might get computed internally by interpolation or PMax).

Brightness adjustment is turned on by default, so definite user action is required to turn it off (at Options > Preferences > Alignment).

The PMax method always makes changes to brightness, contrast, color, and noise as an unavoidable side effect of the way it works. It is the unfortunately flip side of PMax being good at handling complex geometry.

Often the best result is obtained by running both DMap and PMax, then combining the results by retouching so as to use the DMap result over most of the frame area, reserving PMax for areas with difficult geometry where DMap introduces its various kinds of halos. This is covered by the two retouching videos linked at http://zerenesystems.com/cms/stacker...tutorialsindex. The retouching brush in Zerene Stacker is specifically optimized for this sort of work, since it understands how to seamlessly paint details between two images that may have different brightness and contrast.

----------

I hope this is helpful. Please let me know if you have other questions about how Zerene Stacker works, or under what conditions one or another product might work better. I'll try to track this thread for a while, but in general the best way of reaching me is to send email to support@zerenesystems.com , or to catch me in the forums at photomacrography.net . Thanks!

--Rik

[DanK]

Rik,

Thanks for taking the time to post all this detail. I've been stacking for a long time (mostly with Zerene), but there is a lot in your post that I didn't know.

Re running both methods: for some time, I retouched entirely from originals, but I finally tried your suggestion of retouching from one composite to the other, and it is a huge time-saver. In the case of flowers, I generally prefer the appearance of DMap stacks, so I now fairly often retouch from PMax to DMap, avoiding all of the time involving in hunting through a stack of 15-20 images.

Dan

[chauncey]

As a point of reference, I have had the pleasure of communicating with Rik soon after downloading
a trial version of Zerene...I have never had such rapid feedback from customer service...never!

[Glenn NK]

As a point of reference, I have had the pleasure of communicating with Rik soon after downloading
a trial version of Zerene...I have never had such rapid feedback from customer service...never!

This is very true - several years ago I was assessing Zerene and another standalone stacking software.

I sent a question to both Zerene (Rik) and the other - Rik replied very quickly - I never did hear from the other person.

I bought Zerene and use it.

One big advantage of Zerene is that it has a larger user base which is a definite advantage whenever problems arise.

Glenn

[rjlittlefield]

Thanks for the feedback and kind words -- I'm glad to hear that the new info was useful, and also very happy that I've had the pleasure of talking with so many of you earlier!

--Rik

[rudi]

.....
I sent a question to both Zerene (Rik) and the other - Rik replied very quickly - .......Glenn

+1; I have nothing heard from the other, so far..

[DanK]

To add to this, I will mention that one of the previous times I sent a request for help to Rik, the problem was actually with my computer, not the software. Rik explained the problem anyway, which led me (finally) to replace an out of date computer, and the problem vanished.

[jcuknz]

I have yet to bother with stacking but with making panoramas I am basically suspicious of any programme and I would suggest an essential part of stacking a moving object is to be able to erase the bits that do not work, diverge from the sharpness one is trying to achieve. I stitch with an editing programme with layers rather than a stitch programme except for the most basic stitch ... quick and nasty = a programme .... quality comes from human control. That is not a universal guideline as often programmes work well and accurately and better than us humans. Of course a lot of human control is the ability to appreciate what one is seeing as one works the various controls.

I have wondered if it would be possible to adjust focus as the camera shoots in burst mode to reduced the interval between frames.

edit ..glad I read back on this page two as if I do stack I will remember Rik's comment about back focusing which is possible for me having a bellows ... thanks

[rjlittlefield]

I have wondered if it would be possible to adjust focus as the camera shoots in burst mode to reduced the interval between frames.

Yes, this should work fine. I have even stacked from video, by extracting individual frames from the video stream. You're likely to end up with more frames by shooting this way, compared with shooting the same total depth with carefully chosen focus steps. But this should not cause problems, aside from increased processing time and possibly a larger number of frames to consider for retouching.

--Rik

[Glenn NK]

For those that really take stacking seriously, this online calculator may be useful:

http://bvdwolf.home.xs4all.nl/main/f...n/dofplan.html

The author Bart van der Wolf is a frequent contributor on Luminous Landscape forums.

Glenn

[rudi]

No, I have not given up on stacking...

Hereby a snout fly (Rhingia), 3 shots handheld, stacked in Zerene, no crop.
Exif: 7D, 100mm macro + tubes, f11, 1/250, iso 100, diffused flash.


Did I missed some flaws ?
C&C welcomed.