Stacking Mush

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

Moderators: rjlittlefield, ChrisR, Chris S., Pau

phero66
Posts: 35
Joined: Tue Oct 31, 2006 4:43 am

Stacking Mush

Post by phero66 »

So I have tried doing a few stacks in Helicon Focus Demo with pretty dismal results. Basically, the image ends up looking like one big blurry pile of semi recognizable shapes, with a sprinkle of very small specks of detail here and there. The individual frames actually looks MUCH better then the final result. Examples forth coming tomorrow.

Can the gods of stacking help such a poor soul as mine? This is what I have tried:

-Half-quarter profile of a small flower
-Tripod, Mirror Lockup – Individual frames checked for sharpness
-50mm reversed on bellows
-.1 and .5 focus rail increments @f4 (.5 seems to be the best so far)
- stack of 40+ images
-Helicon Focus set to 0,0 for parameters, and also tried at defaults

I figured there would be some errors in processing the files, but what I’m getting looks really really bad compared to the detail I see in the individual frames. I could *fix* the images, but at this stage I would have to paint back in the detail from every frame.

What am I doing wrong?

rjlittlefield
Site Admin
Posts: 23561
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Re: Stacking Mush

Post by rjlittlefield »

phero66 wrote:What am I doing wrong?
From the description, I don't have a clue. Post samples and maybe we can figure it out.

As I recall, Helicon Focus comes with a few small demo stacks. Does your copy process its own demos correctly?

--Rik

phero66
Posts: 35
Joined: Tue Oct 31, 2006 4:43 am

Post by phero66 »

Ok sorry again ran out of time to post pics before heading to work tonight.

But I think I know what the problem is now: focus increments + aperture.

*The Helicon Focus samples worked fine so I know its me*

Test 3:
F8 @ .05mm = better results at default settings, though increasing the radius up to 15 helps a lot, still parts of the image that should be in focus are not. Running radius and smoothing at 0,0 gets the worst results, everything looks blurry with pieces of detail here and there.

I'm thinking that if I have to increase the radius to get better results then there isn't enough transition detail between images for HF to make a good transition?

rjlittlefield
Site Admin
Posts: 23561
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

John,

I usually get best results with default parameters or close to it. I have never seen (0,0) do anything useful -- it causes all decisions to be very local.

Try this... Set default parameters and process the stack. If it's bad, then post out the result and two adjacent frames from the middle.

I'll bet there's something "obvious" going on -- I just have no idea what it is, and may have to study for a while to see it! 8) [*]

--Rik

[*] In the middle of his lecture, the math professor gestures to what he has just written on the blackboard, saying "Now, it's obvious that..." He stops, cocks his head, and looks hard at the blackboard. Then he cocks his head the other way, assumes a troubled expression, and pulls his beard. After a couple of minutes, he excuses himself and leaves the room. Ten minutes later he returns, goes back to the blackboard, and takes up where he left off. "Yes, it is obvious..."

Have I mentioned that once in a while I teach math? :lol:

phero66
Posts: 35
Joined: Tue Oct 31, 2006 4:43 am

Post by phero66 »

Ok so heres the image from test 3, is this what most images look like before having to go in and retouch, or are there other settings I should be trying first thanks!: *i'm thinking the center softness is because the stack wasn't deep enough, though its not putting eveything in*

Not sure what you mean by adjacent frames so the link below should get you a bigger image to review.

Image

A bigger version can be had here:
http://www.artbotphotography.com/200701 ... =8_S=4.jpg

And medium res. source files here (9mb zip) no sharpening after resizing though they could use it):

http://www.artbotphotography.com/Med.zip

rjlittlefield
Site Admin
Posts: 23561
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

John,

You have an interesting and vexing problem that I have not seen before.

Looking at your stack, I can see by eye that the center area of the flower is sharply focused in numerous frames in the middle of the stack, around your _MG_0687. But this area is rendered by Helicon Focus as a blurry mush. (I like your title.)

The only thing I notice that seems unusual about this stack is that detail in the mis-rendered area is both low contrast and not very sharp -- both of these because of the nature of the flower, not your technique.

I'm speculating, but perhaps it's this combination that HF does not handle well.

As an experiment, I tried running your stack of medium resolution images through CombineZ5.

After some arcane tuning[*][/size], this is what it produced.

Image

I have to say, I'm impressed by your bad luck at running into this problem on your very first outing!

From here, I can suggest several possibilities. First, you might try sharpening your images before running HF and see if that matters. Second, you might try a different subject -- one with sharper, more contrasty detail -- to get some (hopefully!) more successful experience. Third, you might try turning this subject into one with sharper, more contrasty detail, by adding some illumination to bring out surface texture in the problematic area.

As I said, this is an interesting and vexing problem. :?

Keep me informed, please!

--Rik

[*] Find Detail(5) vs default (25), turn off filtering after interpolated output.[/size]

rjlittlefield
Site Admin
Posts: 23561
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Of course there are least two more directions to go from here.

Fourth, you might ask Helicon Focus and offer them the stack. Chances are that either they already know how to handle this problem, or they'll be interested to figure it out and add your stack into their test & development suite. And fifth, for this particular stack, it appears not much trouble to use the manual editing functions to fix the stacking errors, although I agree that the process is annoying since this looks like a situation that the computer should be able to handle.

--Rik

Charles Krebs
Posts: 5865
Joined: Tue Aug 01, 2006 8:02 pm
Location: Issaquah, WA USA
Contact:

Post by Charles Krebs »

John... ran your stack through a small amount of unsharp mask, and then through Helicon with settings of R17, S4. Looked OK.. but I have never used R17 before! Must be beginners (un)luck.

Image

phero66
Posts: 35
Joined: Tue Oct 31, 2006 4:43 am

Post by phero66 »

Ok great, *phew* (wipes the sweat off his brow)... Here I was thinking you guys must be doing HOURS of editing to get back the lost details in your amazing stacks. :)

I will run USM and increase the contrast on the orginals before running them through HF again. If that doesn't work I'll send it to HF too and see what they say.

Thanks for the help!

-John

PS. Rik thanks for the example in CZ, I have been meaning to try it but havn't had the time to learn it yet.

PPS. Charles, I doubt it was luck, I suspect magic, or at least super human powers. :shock:

Charles Krebs
Posts: 5865
Joined: Tue Aug 01, 2006 8:02 pm
Location: Issaquah, WA USA
Contact:

Post by Charles Krebs »

John...
You might also try shooting the stack at f5.6. I suspect that what you might gain in contrast and slightly more depth per frame would stack easier. I doubt you would notice any resolution difference.

phero66
Posts: 35
Joined: Tue Oct 31, 2006 4:43 am

Post by phero66 »

Charles and Rik,

Ok I tried implementing your suggestions on a new series since there was a number of problems I found with the first shot (ie: no lens shade, and areas of missing focus because too shallow DOF).

Except for the better DOF, I can't say its processing any better then before. Its very strange, it seems like the program is washing out the shadow areas with the "unfocused" parts of the earlier frames. I'm processing from the back forward, which I think is the proper order...

To get around it I spent about 2 hours painting back in lost data, which brings up another problem = Is there a way to save a project in HF so that you can go back and edit the frames later? When I go to save all I get are standard outputs, and opening one of those does not bring up the edit tab (or whatever its called).

Again I'm at work w/o the new image, but a early version can be had in the gallery: http://www.photomacrography.net/forum/v ... php?t=1711

New setup:

F5.6 @ 3/8mm
Contrast Adjustment before stacking
HF: 17/2
HF Edit: 2 hours re-editing back in the lost detail

Thank you for all your help.

rjlittlefield
Site Admin
Posts: 23561
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

John,

I'm saddened to hear that you're still having problems.

For what it's worth, I've come to think of these stacking programs (and most other software) as being like trained animals. Most of the time they do something close to what you want. But it's seldom exactly right. Sometimes they do something completely weird, for no obvious reason. And if you want them to do something different, the retraining process can be long and uncertain. The best way to get good results is to figure out what the animal wants, and give it to them. <wry grin>

What HF really wants is small, crisp detail at a contrast level that's well above noise.

If by "Contrast Adjustment", you mean levels & curves, that won't help much because they don't change the detail-to-noise ratio.

What's needed is more like unsharp mask with a radius and amount chosen to bring out detail in the subject while not affecting the noise too much. With the reduced size test stack you posted, I got encouraging results (very much like Charlie's) using radius 2 pixels at 50%.

Front-to-back versus back-to-front makes no difference except for the size of image you get. Once I even scrambled a bunch of file names to randomize the order of processing. (The stack was already aligned so I could turn off auto-adjustment.) Literally all the randomization did was to change the bottom bit in some pixels. I attributed that to just arithmetic rounding.

About saving the project so that you can go back and edit -- nope, I don't know any way to do that. If you find one, please let me know too. Seems like I'm always finding something I'd like to tweak.

--Rik

PS. One more hint is to shoot at the very lowest ISO you can afford. That reduces the noise & makes it easier for stacking software to pick out detail in the subject.

phero66
Posts: 35
Joined: Tue Oct 31, 2006 4:43 am

Post by phero66 »

Rik,

You got me eager to return and run another test through with usm instead of lvls. I didn't try the later because once I ran the stack I started mucking with it, and well, you know, you can't go back and edit later...argh! So I just finished what I started and then ran out of time to test again.

-John

MacroLuv
Posts: 1964
Joined: Mon Aug 28, 2006 2:36 pm
Location: Croatia

Post by MacroLuv »

In spite of all troubles I must admit that photograph appearance is quite attractive. :D
Must give a try to stacking some day. 8)
The meaning of beauty is in sharing with others.

P.S.
Noticing of my "a" and "the" and other grammar
errors are welcome. :D

phero66
Posts: 35
Joined: Tue Oct 31, 2006 4:43 am

Post by phero66 »

Well my motherboard crashed this weekend so I could not run any more tests. But I did get some good news from Helicon Focus. It turns out they are planning on implementing some sort of individual frame output feature so that one could work on the stacked image within photoshop and then add in each corrected frame as layers – thus getting the ability to paint back in detail AND save.

Thanks for the compliments Macroluv!

-John

Post Reply Previous topicNext topic