Photoshop is really bad at stitching
Moderators: rjlittlefield, ChrisR, Chris S., Pau
Photoshop is really bad at stitching
Just for fun I loaded my latest stack and stitch into Photoshop to see how it would cope.
So far, there are 230 tiles covering this beetle. Once finished it will be a ~600 megapixel image.
The alignment took just over 90min on a 2013 MacPro with 6-cores, 16GB of ram and a 2TB scratch drive. Photo alignment is a single core process in Photoshop CC so cheaper, but newer I7 would have done it faster. The scratch drive is just an old empty USB3 hard drive so not to use the internal 500GB SSD.
Here are the results:
Yes, this is a 10GB image
As you can see its bad, really, really bad. Almost comically bad.
Lightroom does a MUCH better job of aligning and blending the images, but has size limitations.
Another week and I should have finished imaging the legs and it will be just the editing left to do.
So far, there are 230 tiles covering this beetle. Once finished it will be a ~600 megapixel image.
The alignment took just over 90min on a 2013 MacPro with 6-cores, 16GB of ram and a 2TB scratch drive. Photo alignment is a single core process in Photoshop CC so cheaper, but newer I7 would have done it faster. The scratch drive is just an old empty USB3 hard drive so not to use the internal 500GB SSD.
Here are the results:
Yes, this is a 10GB image
As you can see its bad, really, really bad. Almost comically bad.
Lightroom does a MUCH better job of aligning and blending the images, but has size limitations.
Another week and I should have finished imaging the legs and it will be just the editing left to do.
Dont get me wrong, i dont mean to discourage you. Actually i evny your patience to put enourmous amount of time that this kind of work requires.
Seeing how much distortion is present in the image and comparing it to my results, i suspect your lens is quite far from telecentric, therefore making this kind of project very challenging. I would dare to say - impossible - for me, as i cannot spend that much time on manually stiching the images.
Can you check your stacking software alignment and scaling parameters log? Just curious.
Seeing how much distortion is present in the image and comparing it to my results, i suspect your lens is quite far from telecentric, therefore making this kind of project very challenging. I would dare to say - impossible - for me, as i cannot spend that much time on manually stiching the images.
Can you check your stacking software alignment and scaling parameters log? Just curious.
- rjlittlefield
- Site Admin
- Posts: 23621
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
perdu34, I confess that I'm curious about your motivation for doing this particular piece of work.
I have seen several people go to a lot of trouble to make huge pixel count stack-and-stitch images of insects.
In some cases the goal is to make wall-sized artwork that can be viewed arbitrarily closely and still look good. I can easily understand that.
But those cases also involve perfect specimens and lots of attention to all the other aspects that make spectacular art.
None of that seems to be the case here, so I'm curious.
I get the vague feeling that maybe you're developing technique that will eventually be used for something else, but I have no idea what that might be.
Is this an aspect that you're able to talk about?
--Rik
I have seen several people go to a lot of trouble to make huge pixel count stack-and-stitch images of insects.
In some cases the goal is to make wall-sized artwork that can be viewed arbitrarily closely and still look good. I can easily understand that.
But those cases also involve perfect specimens and lots of attention to all the other aspects that make spectacular art.
None of that seems to be the case here, so I'm curious.
I get the vague feeling that maybe you're developing technique that will eventually be used for something else, but I have no idea what that might be.
Is this an aspect that you're able to talk about?
--Rik
My comment was coming from my experience.
I've used Mp-E65 on FF to get images like this:
Those are 500 megapixels (analyzing software limit) panoramas of prepared samples. About 15 fields of 10 slices deep stacks. Automatically aligned images in photoshop resulted in 3 "sub" panoramas, that i had to manually stich while using warp distortion.
Only after producing final image, i've noticed that it's full of small errors like this:
top of image
bottom of image
I've stacked just ~10 images per FOV of very flat subject. Insect like specimens i wouldnt attempt anything bigger than 2x2 FOV. Maybe 3x3 if i could find a good reason to spend time on it (im often printing 150cm x 100cm 150 ppi from single A7RII FOV and find results quite satisfactionary).
Hence my recent interest with telecentric lenses, that i struggle to understand. Maybe it's a good idea for you too.
I've used Mp-E65 on FF to get images like this:
Those are 500 megapixels (analyzing software limit) panoramas of prepared samples. About 15 fields of 10 slices deep stacks. Automatically aligned images in photoshop resulted in 3 "sub" panoramas, that i had to manually stich while using warp distortion.
Only after producing final image, i've noticed that it's full of small errors like this:
top of image
bottom of image
I've stacked just ~10 images per FOV of very flat subject. Insect like specimens i wouldnt attempt anything bigger than 2x2 FOV. Maybe 3x3 if i could find a good reason to spend time on it (im often printing 150cm x 100cm 150 ppi from single A7RII FOV and find results quite satisfactionary).
Hence my recent interest with telecentric lenses, that i struggle to understand. Maybe it's a good idea for you too.
I've done this too. In my case, it was expending infeasible amounts of time and effort to produce a 30k-pixel square image of a 2mm diameter diatom arrangement I'd made. Stack and stitch with a 10x Plan APO objective. Sure, I could zoom in and see striae and even punctae on individual diatoms. But in showing it to the world, I posted a 1k square image first, followed by an A1 print for an exhibition (an exception, not what I ever planned to do).
In both cases, I put FAR too much effort into producing the original image in comparison to the output resolution that was actually needed.
But! Big but! It was still a very useful learning experience. Now I tend to consider what I'm going to do with the image in order to "balance" the effort put into capturing it. Not what I *may* do - what I *will* do. The only reason I do that now though is because I learned "the hard way" for myself, not because I listened to those who pointed out my wasted effort.
I wouldn't have it any other way. Lessons experienced are infinitely more effective than lessons dictated IMO! With the former, you can incorporate your own desires and proclivities - and sometimes produce something far beyond what is considered possible by accepted wisdom. Go for it!
In both cases, I put FAR too much effort into producing the original image in comparison to the output resolution that was actually needed.
But! Big but! It was still a very useful learning experience. Now I tend to consider what I'm going to do with the image in order to "balance" the effort put into capturing it. Not what I *may* do - what I *will* do. The only reason I do that now though is because I learned "the hard way" for myself, not because I listened to those who pointed out my wasted effort.
I wouldn't have it any other way. Lessons experienced are infinitely more effective than lessons dictated IMO! With the former, you can incorporate your own desires and proclivities - and sometimes produce something far beyond what is considered possible by accepted wisdom. Go for it!
Orthographic stitches are almost impossible to do without either distorting the subject or having stitching artifacts.
In my opinion, the only good methods to make large panoramas are by rotating the lens/camera around the entrance pupil or using a telecentric lens.
That being said, try Autopano Pro or Microsoft ICE for orthographic stitching. Both will be significantly better than Photoshop.
In my opinion, the only good methods to make large panoramas are by rotating the lens/camera around the entrance pupil or using a telecentric lens.
That being said, try Autopano Pro or Microsoft ICE for orthographic stitching. Both will be significantly better than Photoshop.
Sad that Autopano is no longer aliveelf wrote:That being said, try Autopano Pro or Microsoft ICE for orthographic stitching. Both will be significantly better than Photoshop.
http://www.kolor.com/
- iconoclastica
- Posts: 487
- Joined: Sat Jun 25, 2016 12:34 pm
- Location: Wageningen, Gelderland
Hmm.. I can see this is probably not what you were after. But don't throw it away, with all its imperfection it's a work of art in itself.Just for fun I loaded my latest stack and stitch into Photoshop to see how it would cope.
So far, there are 230 tiles covering this beetle. Once finished it will be a ~600 megapixel image.
--- felix filicis ---
JohnyM wrote:My comment was coming from my experience.
I've used Mp-E65 on FF to get images like this:
Those are 500 megapixels (photo editing software limit) panoramas of prepared samples. About 15 fields of 10 slices deep stacks. Automatically aligned images in photoshop resulted in 3 "sub" panoramas, that i had to manually stich while using warp distortion.
Only after producing final image, i've noticed that it's full of small errors like this:
top of image
bottom of image
I've stacked just ~10 images per FOV of very flat subject. Insect like specimens i wouldnt attempt anything bigger than 2x2 FOV. Maybe 3x3 if i could find a good reason to spend time on it (im often printing 150cm x 100cm 150 ppi from single A7RII FOV and find results quite satisfactionary).
Hence my recent interest with telecentric lenses, that i struggle to understand. Maybe it's a good idea for you too.
Maybe you should try another software. GIMP for example
I love photography