Parallax errors?

[shawngibson]

Do I need to worry about parallax errors with a compound microscope when stitching many images? Haven't tried to do this yet. And I don't think Manfrotto makes mini-me versions of it's 360 degree heads lol.

[mrsonchus]

Not really, when stitching, maybe a little more difficult when stacking however.

[gekko]

Do I need to worry about parallax errors with a compound microscope when stitching many images? Haven't tried to do this yet. And I don't think Manfrotto makes mini-me versions of it's 360 degree heads lol.

What do you mean by parallax errors exactly? As far as I know, all you need to do is to make sure that there is sufficient overlap between adjacent sections.

[shawngibson]

Sorry for the late reply.

Coming from the world of landscape stitching. You need to set the camera up properly using a 360 head that is properly calibrated:

Her's an example of the sort of calibration I used to have to do:

http://www.johnhpanos.com/epcalib.htm

So let's say I want to take a 5x5 grid with each image overlapped by 1/3 (standard procedure), and then circle back to the first location (grid 0, 0), refocus (to stack) and go again through the grid, etc. as many focus stacks as necessary. Let's say I do 5 stacks, that would be 5 stitched images (I use Photoshop or NIK for this) that get brought into Helicon. Am I making sense lol?

It is in such a use case that I am worried about parallax error.

Shawn

[gekko]

Someone more experienced that I may give a better answer, but my suggestion would be to take, at each "location", multiple images for a stack, then move to the next (partially overlapping) area, take another stack, and so on. Then stack each location separately with your software, and afterwards stitch the stacked images. Your stacks need not be at identical focus planes in the different locations, and in fact, may cover different ranges, depending on the situation in each location. Again, that is what I think, but others may have different (and better) advice. I don't think that the amount and location of overlap has to be exactly 30%-- I guess that is just a rule of thumb to provide adequate information to the stitching software.

[shawngibson]

Interesting. Definitely worth experimenting here. I know with a camera doing something similar but for the purposes of expanded dynamic range (i.e. same image at different exposures) I would work the grid, then go back and change the exposure and start over. But with my system, I had full locks, specific degrees, etc., so it was very precise (something I've lost on the microscope). This thing:

http://www.manfrotto.com/multi-row-panoramic-head

Without those locks, I most likely wouldn't be framing exactly the same image each pass, so your idea makes a lot of sense to me. I can also go back and look at the first image of each grid section, to see where it was focused, and focus on the same plane using your method. This just might be the way to go:)

[gekko]

The reason for my suggestion, as you imply, is that for stacking, you need to use the same field of view, and it is much easier to do this by taking all the images for the stack of one part, then move to a different part, and repeat. So you will stitch the stacked images. On the other hand, you could try taking at one focus point multiple overlapping images, then repeat that for a different focus point, etc., then stitch the images at each focus point, then stack the stitched images. I don't think that would require "indexing" in the sense of always coming back to exactly the same field of view that you had taken earlier, since the stitched images will be "complete" at each focus level. However, the former method is more intuitive to my way of thinking. My 2 cents' worth.

[shawngibson]

I'll try it your way tonight, quick and dirty, as an experiment: 3 stacks, 5 horizontal overlapping stitches.

[mrsonchus]

Hi, I'm no expert but have indeed used Gekko's suggested method with acceptable success, I'd go to that one first personally....

[shawngibson]

Well I just did 2 tests, my way and Gekko's way. My way, it was very comfy, what I'm used to. Gekko's way, I had to write down my steps in Notepad and follow for each image lol.

I've been very lazy with all of these early images, so I've been putting them all on my desktop, as opposed to one of the many drives on my small network.

I opened the first test, my way, and got nowhere in Photoshop (just kept spinning). So I finally said heck just upgrade to the CC version, and pay $22/month. Started that process. Downloading now. Also decided to move all my desktop photos to a proper drive, but when it was done, my second test of tonight isn't there! It's just...gone...

Maybe it will show up when PS finishes installing, or when I restart...or something. Grr

[shawngibson]

Photoshop is back up. NIK is dead. Here's what I was able to accomplish, not great. Approx. 5 x 5 and 5 layers deep. Was very challenging...and not great results.

[shawngibson]

Better?

[mrsonchus]

Wow -they are spectacular - good results I'd say!

[gekko]

I fully agree with John. Both sets are very nice indeed. Since the stacking and stitching has already been done, would you care to post a (probably too boring ) picture in "natural" color? I am sure that it would be equally wonderful.

[shawngibson]

Thanks guys. I'll post the original stack/stitch when I get home tonight.

I assume this 14mp Omax USB camera is probably a point-and-shoot-sized sensor, so a 5x5 stitch after overlap is probably still smaller than an APS-C

(This might not seem a problem when viewing on a screen, but one of my goals here is to create very large, approx 4'x4', oil/wax paintings of some of these images...so I need all the sensor resolution, lens resolving power, dynamic range, etc, that I can get.)

Shawn

[gekko]

To determine the minimum sensor resolution needed to make use of the resolution that your microscope optics are capable of, given a certain sensor size, you can plug in your numbers in the spread sheet developed by Charles Krebs:
http://krebsmicro.com/relayDSLR/relay_micro.xls

[shawngibson]

Awesome, thanks Gekko. This will be very helpful to help me match a camera/objective upgrade.

As promised, here is one of the original files for the above. The prepped slides are stained, 2 red and 2 green fleas. Unfortunately this whole kit of 20 or so metazoans is all green, red, or very dark. I don't get the point. Shouldn't stains target only certain types of cells/molecules? Ah, or maybe it's for contrast...yeah that might make sense.

Shawn

[gekko]

Thanks Shawn. I like your 2nd set ("Better?") most of all, although they are all very nice. As to your question about staining, I defer to our resident expert, mrsonchus, but in my ignorance, I suspect that those kinds of prepared slide are heavily stained to show the entire organism, not to differentiate different organs within them.

[shawngibson]

I prefer "Better" too, as the green and red on white looks like...green and red on white lol. The other one has something more touchy-feely, I think.

On the other hand, all else being equal, this shot, with polarized light and unstained specimens, may look awesome:)

[mrsonchus]

Hi Shawn, those slides are really very nice - the single-colour stains do what I personally believe (from my experience although amateur and very limited..) to be the single most important thing, give plenty of contrast. This makes virtually all observation far more efficient and just plain easier, both on the eye and the camera-sensor.
The differentiated-stains (metachromatic stains do this automatically - try the superb 'TBO - Toluidine blue for almost everything..) picking out with colour differences in addition to contrast enhancement, the various cells, tissues, organs, chemistry etc are even better for most purposes I think, and they are always stunning and impressive visually.
You can't go wrong with good contrast I find, more expertise is needed to interpret what is seen with only 1 colour of course, and I often find myself spending hours interpreting an image in terms of the above - that's one of my favourite activities during the pursuit of which I learn (or try to at least ) a lot.
I love studying detailed images, with labels (drawings are simply better than photographs for this as they are by design able to emphasize information for this very purpose) and am currently doing just that with my Sonchus flower-head and ovary slides.

To glean maximum details (post-processing is usually very valuable for this) from an image make sure when you capture the image that the 'ends' of your exposure are well within the limits of your camera's capabilities - easily verified my checking that the images's histogram isn't 'overflowing' at either end (i.e. dark and light extremes - L & R ends). Remember, once that envelope is breached either with pure black or pure white any details contained therein are irretrievably lost. Make your shadows as dark as you can within these limits and your bright areas as bright as you can (i.e. 'fill the histogram) and you will capture optimum distinguishable detail - a good contrasty stain will greatly facilitate this, and your slides have such a stain!

I often use a facility within PSE-9 that, similar to but not quite the same as the usual 'brightness and contrast' adjustment, lifts the shadows and can lower the bright areas to your preference, basically optimizing detail and therefore information contained within your image as exposed.

Very nice images, lots to interpret - thanks for posting them.

[mrsonchus]

I think I've teased-out a bit of extra detail contained within your excellent images with trusty PSE-9;
Hope it's OK to meddle with your images..

Hidden gems..

enhanced_shawns_fleas.jpg (44.48 KiB) Viewed 9763 times

Sorry to mess-up your images, couldn't resist it!

[shawngibson]

That's the best out of the bunch, John. The depth is awesome.

I have so many threads here (trying to learn) that I'm not sure where to say what. I get a lot of advice by you, Gekko, 75RR, APO, and others. I read it all, but sometimes I'm like "this is for tomorrow" and I bookmark it for later, because I have so much to learn; and learning usually comes in small steps.

I only had 2000ms to deal with on the exposure slider, so I had to crank the gain; I would never do that normally. To me, it's like shooting at ISO 3200; I'd never do that, but I had to respect Toupview's limits (slider only goes to 2 seconds on the slider; maybe I can punch in a bigger number somehow and keep the gain at stock?).

Shawn

[shawngibson]

Can you post a bigger version of that, John? It's awesome. I can send you my original PS if you need it...

Shawn

[shawngibson]

You got a lot of detail in the abdomen of the left flea, that I couldn't coax out. Bravo!

[gekko]

John, I fully agree with Shawn: awesome!

[shawngibson]

John's image is extremely beautiful, eh? And has a lot of organ detail.

[gekko]

John's image is extremely beautiful, eh? And has a lot of organ detail.

Well, it was your image. John used some of his magical manipulations to bring out what was already there .

[shawngibson]

OK *our image

[mrsonchus]

OK *our image

It's all in the source-image I'd say.

[mrsonchus]

Can you post a bigger version of that, John? It's awesome. I can send you my original PS if you need it...

Shawn

I haven't a larger version, but I'll have another go at doing it again, I'm embarrassed to admit that I didn't note down my steps! I've got to practice-what-I-preach I think!

Back soon I hope...