howto improve stacking quality

[lolhonk]

Hi,

i made this stack from a tiny fruitfly (<0,8 mm) out of 120 images, with stepsize of 5 or 10 µm (im not sure anymore ) with my lomo 8,7x @fullframe sensor. Lighting: flash

the image is 1920x1280

http://i.imgur.com/xR7J9V9.jpg

I used zerene stacker with standard settings for PMax. Is it possible to improve the structures like the hairs and the ocelli and the tiny hairs between the eyes? And some transition f.e. from head to abdomen or in the mouth, nose part are not so good.

I tried to use the Dmap with different radius but it didnt get any better. Are my template images bad? Stepsize is to big, to small ?! Or is the problem the resolution of the lomo?

im glad to get some advices from you!

[lolhonk]

Hi,

do you need more information? Or is the only way to improve it to use retouching and or/ better optics?

[ChrisR]

Hi
The first think I noticed was that it's heavily sharpened. If it's that obvious, then it's arguably too strong!

For best advice, if you take the Pmax image, don't sharpen - "Curves" if you need to would be ok but tell us, then crop a 1024 square out of the middle and post that on the forum, so we can see "actual pixels". Use enough jpeg compression to just make it less than 300kB for the forum.


To make the transitions more gradual, you can go one of a few ways:

One is to make the stack deeper than you want, then apply some blur to the source images towards the far end of the stack. Those images can be taken at larger steps than the rest, if you're going to blur then anyway.

Second is to use an adjustable iris on the objective. They do exist for rms objectives , but only (cheaply?) on the used market. (M42 are available new). I once bought a "funny looking" objective and it had an iris adapter on it, rms both sides. An alternative would be to make a slide-on black paper cover with a small hole in it, to put subject side of the objective. I tried it before I got an iris - it can work ok but can also give vignetting.
The idea then is to use the smaller aperture for a frame or two, at the end(s) of the stack. You will usually have to do some retouching, to remove double imaging etc from the sharp parts of your output image. I've used 2-3 stops smaller than the maximum aperture - which of course means needing 4 or 8 times the exposure. I usually use flash, which can be turned up.

[rjlittlefield]

As ChrisR says, we need to see images that have been processed less.

We also need to know more precisely what you're thinking about when you say things like "improve" and "not so good". When I look at your image, the tiny hairs between the eyes look fine except for diffraction and over-sharpening, while the big bristles suffer from the "transparent foreground" effect that is discussed in the first video tutorial about retouching, and again in http://www.zerenesystems.com/cms/stacke ... foreground.

--Rik

[lolhonk]

@chris & rik

thank you very much! I made new images without sharpening etc. output direct from zerene stacker.

image is stacked from 37 images with stepsize 60 µm. Sony a7rii with lomo 3,7x.

i resized it to 1024px and i had to compress it to reach under 300kb:


100% images




- sharpness get lost in direction to the background (see last image)
- it looks "washed out" f.e. look at the eye @ image 2
- or did i want to much?
- i tried the same with half the stepsize -> 30 µm but it didnt get better.

best,

[rjlittlefield]

Thank you for the further information.

Your first item, "sharpness get lost in direction to the background", looks like a typical DMap artifact that we usually call "loss-of-detail halo".

That happens because the depth map cannot slew quickly enough from background to bristle and back again. Instead, the depth map moves between background and bristle over a span of several pixels. Where it does that, the image will show source frames that are out of focus. The halo consists of pixels chosen from those out of focus frames.

The best available cure for loss-of-detail halos is to use the PMax stacking method. PMax has no trouble with instantly switching from background to bristle and back again.

But there's a tradeoff. PMax also accumulates noise and alters contrasts and colors.

So, for subjects like this one, you may have to choose between preserving the maximum amount of detail (using PMax) and retaining exact colors (using DMap).

Often the best compromise is to run both PMax and DMap, then use retouching to combine the best parts of both.

For your second item, "it looks "washed out" f.e. look at the eye @ image 2", I cannot tell exactly what's going on. What I see in the image looks like normal appearance of a partially dried out eye, where the dark interior has separated from the clear cuticle, producing an overall grayish effect. I suggest to check your source images and see if they show something different.

Also, you mentioned:"i tried the same with half the stepsize -> 30 µm but it didnt get better." I believe the Lomo 3.7X objective is NA 0.11. If that's correct, then its nominal DOF is about 45 µm. That's for green light; it will be less for blue. For subjects like this, I would recommend using no more than 45 µm, safer would be 30, but I can easily believe that the problems you've identified would not be changed by smaller step.

--Rik

[abpho]

Those are looking good Lolhonk. You are on the right track.

[austrokiwi1]

Abpho. are you capturing in raw or Jpeg? It looks to me like you may have been using Jpeg.

[abpho]

Are you sure you mean me austrokiwi1? I am not the OP.

[austrokiwi1]

My mistake , I should have referred to LOLhonk

[lolhonk]

Hi guys,

thank you for your advice! here is my next try:

As Rik mentioned i made a combination of pmax and dmap and i was a little bit shocked about the tiny DOF of the lomo (45 µm) thats why i set the stepsize to 20 µm which resulted in 109 images.

I shoot always in raw (no eletronic shutter because i need flash ) And export it as tiff to zerene stacker (This takes years and the stacking takes decades and yes i have a good computer, but thats the penalty of the 42 MP sensor)

Image 1: mix from pmax & dmap ~ 30 minutes retouching, no sharpening etc. just resized to the ugly 1024px and compressed to under 300kb :-/

for better quality pls look here: http://imgur.com/a/LYz1a



Image 2: i tried to get the best from pmax to avoid the "loss-of-detail halo" but still the parts in near to the background are not as good as the other parts.



Image 3: Problems are now in the darker parts, i dont get enough light here



Image 4: I was surprised that the wings and the edges of it are very good. I t feared problems here because they are so transparent



@Chris or other: what does Chris mean with this:

The idea then is to use the smaller aperture for a frame or two, at the end(s) of the stack. You will usually have to do some retouching, to remove double imaging etc from the sharp parts of your output image. I've used 2-3 stops smaller than the maximum aperture

Whats the benefit of this? The resolution get worse but the edges are better?

I noticed that the position of the object has a great influence of the quality of the stack. And this cuckoo wasp (btw a terrible name of this beautiful insect, in german we call it: goldwasp) isnt easy to stack because the billions of reflections and tiny structures which get lost in the resulting halo of this reflected light.

I think its better then my last try. What do u mean?

best,

[austrokiwi1]

I know what you mean about exporting to tiff taking ages! Are you using capture one pro as the raw editor? if you are I was wondering if you have been using the halo suppression tool? I would have thought that tool would help with sharpening

[rjlittlefield]

Image 2: i tried to get the best from pmax to avoid the "loss-of-detail halo" but still the parts in near to the background are not as good as the other parts.

Bristles on the edges are almost always low contrast because they have been contaminated by out-of-focus blurs from closer parts. See http://www.photomacrography.net/forum/v ... 042#135042 for more details about how this occurs. The diagram there uses a sphere in the foreground, but the principle is the same. When you are focused on the parts in back, the view is partially blocked so that only one side of the objective, sometimes even less, is actually seeing what you are interested in. The rest of the objective is seeing OOF foreground parts. The problem is particularly bad in dark areas of the background parts. That's because dark is more easily contaminated by light, than light is contaminated by dark.

Problems are now in the darker parts, i dont get enough light here

The bright blobs that you have marked with red arrows are almost certainly DMap artifacts caused by having the slider a little too far to the left.

What happens is that Zerene Stacker mistakes the spreading edges of out-of-focus blurs as detail to be preserved. Wherever that "false detail" provided by the OOF blurs has more contrast than whatever real detail is present in the focused frames in that same area, DMap is inclined to show the blurred frames rather than the focused ones. If it's allowed to do that, then the rendered image will switch quickly between focused frames showing correct color, and unfocused frames whose colors are contaminated by surrounding areas, bluish-green in this case. In the case you're showing here, those solid black areas have essentially no real detail of their own, so they are easily contaminated by the spreading edges of OOF blurs.

The key to getting a clean rendering in this case is to set the contrast threshold slider high enough that all of those bogus blotches get covered up by the black-in-preview mask. Essentially you're telling the software that those areas are such low contrast that depth-from-detail is ambiguous, so those areas should be ignored when computing the depth map. As always, areas that are covered by the black-in-preview mask will have their depth inferred from surrounding areas that have strong enough real detail to be handled correctly.

I was surprised that the wings and the edges of it are very good. I feared problems here because they are so transparent

Sometimes DMap will have trouble with transparent areas, because DMap is never able to show detail from two very different depths in a small neighborhood. PMax generally does a good job in this situation.

I think its better then my last try.

I agree that this later one is better than the earlier.

--Rik

[anvancy]

I was searching for information like this. Thanks lolhonk for this thread.

I have another question. What are the advantages of putting TIFF images inside Zerene? My workflow involves, take raw images in LR, base process for exposure curves corrections. Export them as JPGs into Zerene and stack them. Export them as TIFF and bring back in LR. Export stacked TIFF from LR as JPG. Edit this in photoshop and present the result.

I have a moderate rig. AMD 1090T+8GB Ram. I use the canon 80D RAW files, so I assume my rig will go slow during slabbing and subsequent stacks?

Anvancy

[rjlittlefield]

What are the advantages of putting TIFF images inside Zerene? My workflow involves, take raw images in LR, base process for exposure curves corrections. Export them as JPGs into Zerene and stack them. Export them as TIFF and bring back in LR. Export stacked TIFF from LR as JPG. Edit this in photoshop and present the result.

The main advantage is that TIFF lets you use 16-bit color instead of just 8-bit color as with JPEG. That preserves finer gradations in color, which gives you more freedom to adjust levels & curves in postprocessing without getting contour bands in areas with smooth gradation such as OOF background. In your current workflow, you are quantizing to 8-bit color in the JPEG stage. Some of that will be preserved in the stacking process and will still be present in the TIFF output, even if you save output as 16-bit TIFF.

Stacking will be somewhat slower with TIFF, but how much slower depends heavily on how your computer is configured (especially I/O speed), and on what Options you have selected in Zerene Stacker (especially overlap computation and I/O). You'll have to try it to see how much it matters.

--Rik