Refining Technique

[PaulFurman]

Polypodium californicum - Polypody Fern

I was interrupted with a computer switch before finishing this a week ago so the specifics are not fresh, like lighting, which Pau solved in a similar shot with polarized fiber optics. I think I had to remove my diffuser plastic hood to squeeze light into this vertically mounted specimen on my horizontal rig in order to maintain reasonable shutter speeds of around 2 seconds.



Oly 20/2 at about 10x on 12MP full frame.

I did the stepping on the rear bellows standard this time which produced a fairly large change in magnification:

___

... but hopefully improved alignment problems that I've had with deeply overlapping features. Probably not much of an issue for this fairly flat subject though.

You can see a slight detail improvement between the upper left and lower right due to magnification change:

__

I shot 225 frames which was too many. It's easy to make small steps using the rear standard; no micrometer involved, just the bellows knob. This version is only 48 frames. I tried stacking the whole thing and there is a slight improvement:

__

... I might use every third shot if I did it again.

Also, I tried DMap with Zerene stacker, with the default contrast level of 74. That came out very clean with a little less highlight blowout and some transitions improved but it put all the background detail near the raised spore cluster out of focus so I stacked with PMax and they evened out nicely just like that. Minimal retouching needed. The background areas were a tad more hazy after the blend (I think), so maybe more work to perfect that with retouching in other situations.

[rjlittlefield]

Nice image, Paul.

A lot of the pictures are showing as "THIS PHOTO IS CURRENTLY UNAVAILABLE". Something about Flickr access?

There are a couple of typos in your description of processing. I'm pretty sure you meant "DMap...with the default contrast level" and "PMax...evened out nicely".

By the way, the contrast level of 74 is not default, it's just remembered from the last time you set it.

If you can figure out about those unavailable images, I'd appreciate it. From the text, it sounds like there's interesting stuff to see and think about.

--Rik

[PaulFurman]

I fixed the images, will check typos later, thanks.

[elf]

I fixed the images, will check typos later, thanks.

Links are still broken

[PaulFurman]

OK I got it all fixed. I mixed up D & P - Map & Max because I don't understand what they abbreviate. I see now why Map is called map, like a topo map as you watch it build zoomed in, the Max one seems more of a blend overlay kind of approach rather than hard cut-off lines. Theoretically, I suspect Map would work better with the bazillion micro-steps in the full set of 224 frames.

BTW the bottom crops also show the subtle improvement between straight Max on the right and retouching in highlights from Map on the left. All the improvements I point to are subtle but they add up.

When setting the contrast for Map, I'm never sure where to put it because the resulting image is gone. I suppose the way to experiment with all this is with just a small number of frames or a small crop so it's faster.

[rjlittlefield]

"DMap" stands for "Depth Map". A single-valued surface, like a topographic map of elevation, is passed through all points where the best-focused frame can be accurately determined. At pixel positions where the best-focused frame cannot be accurately determined straight from the image data, depths are "guessed" by interpolating from nearby points where it could be determined. Final pixel values are determined as a weighted average of two frames, one on either side of the depth map.

"PMax" stands for "Pyramid Maximum". The key difference is that instead of working with pixel values directly, PMax decomposes the image into a bunch of different levels of detail, operates on the different levels independently, then puts the results back together again. The word "pyramid" comes from thinking about a pyramid with lots of fine detail at the bottom, and a little coarse detail at the top.

Behaviorally, PMax is kind of like using DMap with big neighborhoods on the coarse detail, while simultaneously using DMap with medium neighborhoods on the medium detail, and likewise on the fine detail. The "blend overlay" you mention is due to recombining data at various sizes of detail.

PMax has no adjustable parameters because a) the neighborhood size automatically adjusts to match what is found in the image, and b) working on the decomposed form removes the need to ignore uncertain data. These are why PMax is so good at finding and preserving detail even when it's not very sharp or is low contrast and swimming in noise. Unfortunately, as subject features go into and out of focus, information sort of "leaks" from one level of detail to another. This is what causes contrast buildup.

When setting the contrast for DMap, I'm never sure where to put it because the resulting image is gone.

As background info... At the stage where you get to pick the threshold, only some intermediate image has been formed. That image consists essentially of a crude DMap result, but with no explicit smoothing and no special handling of uncertain data. When the slider is full left (Percentile: 0.0), you're seeing all the pixels. As the slider moves right, pixels corresponding to uncertain data are dropped out of the display, becoming rendered as black. Pixels rendered as black are temporarily gone from the display, but can be made visible again by moving the slider left. All of this work is usually done on an image that has been reduced by 2X from the actual stack resolution, mostly to improve response time for the interaction. It is only after the threshold has been determined that the final depth map is constructed, and after that, another pass through the stack is made to accumulate final pixel values.

Given this background, I don't understand what you mean by "resulting image is gone."

Are you talking about pixels being blackened, or that the 2X reduction causes fine detail to be not visible while you're setting the threshold, or that what you see in the preview does not match what DMap produces as final output, or something else?

--Rik

[PaulFurman]

Thanks Rik, that helps a lot.
I'll have to experiment more with smaller chunks.

What I meant by 'the image is gone' is: I recall seeing the stack as it builds and spotting the trouble areas but once it blinks into threshold decision mode, I can't see the problems any more, just an abstraction. Maybe my memory or HD space is running out and the rendered part goes black rather than showing as an image. Sometimes that happens when retouching: it'll go black till I paint in a spot then reappear but often the brush leaves a square black trail. My recollection of the DMap issue is vague as it's just that one point of decision (and confusion), then I choose something randomly or from memory and it's over.

[rjlittlefield]

Sometimes that happens when retouching: it'll go black till I paint in a spot then reappear but often the brush leaves a square black trail.

I haven't heard of this problem before. I presume the black trail goes away spontaneously sometime later? This sounds like one of the many "repainting" issues that turn out to be specific to a particular graphics card, driver, and acceleration settings. Often a good first step is to go into Display Properties > Settings tab > Advanced button > Troubleshoot tab, and slide the "Hardware acceleration" slider all the way left to "None". If that takes care of the problem, then gradually move the slider as far right as you can.

--Rik

[PaulFurman]

I finally got back to this. In DMap, with 5 frames of 12MP, when setting the threshold, the area to be ignored shows as black. That black area increases with threshold number but I can't see what it's going to be replaced with, presumably the first or last frame. For the set I just tested, a threshold of zero seemed fine so maybe not meaningful.

Unfortunately my laptop video card doesn't allow me to adjust video acceleration. The video card is what fried when my computer died last month. I can work around the repainting trails usually, and that varies depending on the size of the project but I've only ever seen black for the DMap threshold adjustment.

[rjlittlefield]

with 5 frames of 12MP, when setting the threshold, the area to be ignored shows as black. That black area increases with threshold number but I can't see what it's going to be replaced with, presumably the first or last frame.

The part about "first or last frame" is not correct. The depth in those areas is estimated as some distance-weighted average of surrounding known depths. Then the pixel values are obtained as usual from whatever frames are indicated by the depth map. The reason you can't see what the result is going to look like is that the calculation takes too long to do interactively. When you finally make a selection of threshold, and the progress bar scrolls for some seconds, that's how long it took just to fill in the depth map. Then to get a final result requires going through the whole stack again. That's the second half of the DMap process.

--Rik