Haematopota pluvialis

AndrewC · Post by **AndrewC** » Sat Jun 20, 2009 3:26 pm

Finally found a Tabanid - it mistakenly chose my car bonnet as a nice place to sun itself and got swiftly collected.

Nothing particularly rare and exciting - I think this is the commonest species in Northern Europe

By the eye spacing I guess it is a female ?

Three images taken in succession with an El_Nikkor 75/4 @ f8, El-Nik 50/2.8 @ f5.6 and then a Nikon PLAN 10x

All stacked with ZS PMax with no touchup. One problem I've been noticing with ZS is that it can lose detail in dark areas like the patches on the face. There is detail there prior to stacking but it gets lost. I'll probably try and reconvert the images tomorrow for the 10x stack and boost those areas prior to stacking to see if I can improve it.

The second picture is my favourite. The third is really included to show the hairy eyeballs (that just can't be comfortable

) and the amazing difference in textures.

rjlittlefield · Post by **rjlittlefield** » Sat Jun 20, 2009 3:47 pm

Nice images, as usual.

AndrewC wrote:One problem I've been noticing with ZS is that it can lose detail in dark areas like the patches on the face. There is detail there prior to stacking but it gets lost. I'll probably try and reconvert the images tomorrow for the 10x stack and boost those areas prior to stacking to see if I can improve it.

You're the first person to mention a problem like this, but I've been expecting it from somebody.

I'm pretty sure this problem is due to contrast enhancement. PMax tends to make the brights brighter and the darks darker. In my testing, it was common for the whites to blow out, so I added some HDR-type code to reduce the frequency & severity at the bright end. But the darks were well enough behaved that I didn't take time to do something similar at the dark end. It sounds like you have some stacks that would be good test cases so I can develop code to handle the dark end too.

In the meantime, rather than modifying the source images, try saving as TIFF with 16 bits RGB and put a checkmark in the "Retain full dynamic range" box. That will make sure that nothing gets clipped, leaving you free to do whatever you need with curves & levels in Photoshop before saving as ordinary 8 bits RGB clipped to [0,255].

--Rik

Harold Gough · Post by **Harold Gough** » Sun Jun 21, 2009 1:17 am

Andrew,

Maybe film would be a better medium?

Harold

AndrewC · Post by **AndrewC** » Sun Jun 21, 2009 5:16 am

Harold Gough wrote:Andrew,

Maybe film would be a better medium?

Harold

If one wanted a final image like this:

Film might have a higher dynamic range on emulsion but you still need to digitise it to stack it ...

AndrewC · Post by **AndrewC** » Sun Jun 21, 2009 5:23 am

So tried working the entire flow in 16bit (converted and stacked files) but I still lose stacked black details. This is the best I could come up with using fill light during conversion and shadow recovery. I'd have to admit that the detail in the blacks is marginal prior to stacking but it was there !

Here's my best effort and a 100% pixel crop from the centre. It's embarrassing how much time I spent trying to remove the hairs visible in the light grey area. They aren't stacking artefacts but are actually shadows and in some cases reflections. Also, check out the entrance pupil of my lens visible in the eye facets !

rjlittlefield · Post by **rjlittlefield** » Sun Jun 21, 2009 11:53 am

AndrewC wrote:So tried working the entire flow in 16bit (converted and stacked files) but I still lose stacked black details.

A good test case, then. If you're willing to send a copy for R&D purposes, I'm sure that support@zerenesystems.com would be happy to hear from you.

It's embarrassing how much time I spent trying to remove the hairs visible in the light grey area. They aren't stacking artefacts but are actually shadows and in some cases reflections.

Hhmm... I suggest leaving the image alone, and using a few words to explain what you think is going on. The shadows & reflections can be confusing, but they can also be informative if the viewer understands how to interpret them. Besides, it's a lot faster. A bit safer too, come to think of it. If you edit out stuff that you think are shadows and reflections, and it turns out they were something else, then just possibly you've changed the nature of the subject.

Also, check out the entrance pupil of my lens visible in the eye facets !

Ah yes -- the entrance pupil, the rings on the front of your lens, and the surrounding diffuser. It occurs to me that there's some analogy to the Heisenberg Principle in these high mag shots -- it's impossible to capture the appearance without changing the appearance!

--Rik

beetleman · Post by **beetleman** » Mon Jun 22, 2009 9:48 am

Great photo Andrew, I started laughing to myself as I wondered what it was like to look through those eyes

I wonder why they evolved those bands and colors

ChrisR · Post by **ChrisR** » Mon Jun 22, 2009 10:38 am

Film might have a higher dynamic range

I didn't think it did? Velvia 50 - lovely but so difficult.
Presumably you're shooting Raw files?
I was astonished how much more information was in the blacks when I started looking at Raws. Would it be worth pre-processing the files, flatter (etc /wo'evva) to suit the Stacker?

Actually I can't see much dynamic range in the black area on the unstacked image either!

AndrewC · Post by **AndrewC** » Mon Jun 22, 2009 12:55 pm

ChrisR wrote:
Film might have a higher dynamic range
I didn't think it did? Velvia 50 - lovely but so difficult.
Presumably you're shooting Raw files?
I was astonished how much more information was in the blacks when I started looking at Raws. Would it be worth pre-processing the files, flatter (etc /wo'evva) to suit the Stacker?

Actually I can't see much dynamic range in the black area on the unstacked image either!

I shoot in RAW but I might switch to jpg for stacks, or buy some Terabyte drives. There really isn't much detail in the blacks, and when you start off with a not a lot and then lose a bit you have even less. I managed to pull a fraction more detail (or perhaps I'm kidding myself) by upping the sharpening and detail extraction at conversion. You've got to give the stacking software something to work on

lothman · Post by **lothman** » Mon Jun 22, 2009 2:01 pm

I put your last pic in Photoshop and IMO could get some details in the black spot. Also the slight overexposed eye can be made crisper

I hope you don't mind editing your pic - otherwise I'll delete it.

edit: pic deleted - next time I'll ask you.

AndrewC · Post by **AndrewC** » Mon Jun 22, 2009 2:08 pm

Well it is normal to ask before editing and reposting someone's image but it doesn't bug me particularly.

Yes, I could have done more in Photoshop but didn't have the inclination, in this particular exercise I was really just exploring what I could do with processing prior to the focus stacking in order to get the detail to enable further tweaks.

Actually the bottom line here is that it was a really marginal image to expose, focus stack and retain detail in all areas. What I would really like to do with something like this is construct an HDR focus stack, but that brings extra challenges related to how the stacking software aligns and scales individual frames differently from the different exposure runs. The same feature might appear to be in a slightly different location when exposure bracketing depending on local contrast, well I think that that is what happens but Rik I'm sure knows better. The solution might be to run a HDR for each focus layer but that would be very time consuming.

Andrew

rjlittlefield · Post by **rjlittlefield** » Mon Jun 22, 2009 11:22 pm

AndrewC wrote:What I would really like to do with something like this is construct an HDR focus stack, but that brings extra challenges related to how the stacking software aligns and scales individual frames differently from the different exposure runs. The same feature might appear to be in a slightly different location when exposure bracketing depending on local contrast, well I think that that is what happens but Rik I'm sure knows better.

That's a good description. What happens is that when you change exposure, local contrasts get bigger in some places and smaller in others. The alignment calculation gives slightly different results depending on which regions of the image have higher contrast. Those slight differences accumulate frame after frame, so with two deep stacks the end results can be significantly different, as shown in your other thread. (For the record, "deep" means 160 frames for the other thread, 189 for this one.)

The solution might be to run a HDR for each focus layer but that would be very time consuming.

That might work, but I'm not sure. The whole HDR scheme is pretty painful anyway, what with tripling the number of input frames.

So I'm a bit more enthusiastic about solving the fundamental problem, whatever that turns out to be.

This stack is different from any of any others I've worked with in that the troublesome area is dark and low contrast when it's focused, but fairly light and noisy in other frames where it's OOF. For example in the JPEGs you sent today, one area is mean 35, stdev=3 when it's focused, but mean 131, stdev=5 when it's maximally OOF and swamped by OOF light surroundings, at the far end of the stack. This is probably causing some interesting interactions with noise accumulation and contrast buildup.

Developing a good solution for this problem is going to take some thinking...

--Rik

Haematopota pluvialis

Haematopota pluvialis

Re: Haematopota pluvialis