C1 22 Pano Stitching Issues
Hi,
I was eager to use the new stitching functionality. This was the only thing missing for me compared to may way old LR 6.14.
As I started to stich first Images taken with a 50mm lens handheld I noticed, that C1 22 did not manage to keep the horizon straight/level:Is there a workaround? Overlap is not the problem here since the part of the image thats being wrong is in one shoot and the overlap to the connencting shot is at least 50%. Second issue is that c1 did only stitch 2 of 5 selected pictures...
LR 6.14 did the same stitch with no hindrance or hesitation with all 5 Images:
Feeling a bit of buyers remorse now because this was one of two reasons for me to step up from c1 20 to ease my workflow by doing it all in one software. As it seems I still have to long for LR 6.14 for panos.
Greetings
Alex
-
I think there is not much you can do, as you can only influence the pano with the lens corrections (beware light falloff) and by selecting the projection method. I don't use LR but my experience with others is that not all software stitch every pano equally well.
I haven't had any problem of alignment in my (limited) beta tests, but my images had more "features".
The beta forum was relatively "unbusy", if I take this as an indicator either the majority of beta testers were very satisfied, or not many users have participated in beta tests. I assume the latter, which means C1 will probably be very happy to receive your sample files in a bug report ("submit a request" on top of this page).
My guess is that C1 will improve the new pano feature in a minor dot release.
0 -
During beta testing, I encountered light fall-off problems at the seam boundaries and HDR merging was less than satisfactory. I processed four different pano sets, made up of five or six images and never saw the issue that you reported; I thought that stitching was fine. BTW, when I shoot panos I use a tripod and manual rotator and put effort into leveling the tripod/camera.
The light fall-off was easy to correct in processing. The poor HDR merger is a different story, hopefully this will be high on the C1 to do list.
0 -
Alex,
You wrote:
"...and the overlap to the connencting shot is at least 50%"
This may be the problem.
When testing I had a result, using 8 images, that was less than ideal and quite similar to your result in the way the problem was presented.
Feedback suggested that fewer images might give a better result and that was indeed the case. I revisited the images and chose the set that provided the same coverage but less overlap and the results are fine.
The underlying message is that the development work to date is seeking to provide a very simple but effective way to stitch images together to create a new "RAW" file that can then be used for the final editing work. But the key message seems to be to keep it simple and expect to avoid making complex alignment and blending choices.
For those that like drilling deep into the technical side of picking alignment points and manipulating blending choices what is currently on offer may be insufficiently comprehensive to satisfy their curiosity. For others, the idea of being able to shoot choose just a few images to get a decent quality instant result with little user input but still have a RAW file to play with may be very appealing.
Sometimes the "hands-on" approach is a necessity and there is no alternative to using a highly interactive application that allows a lot of control choices. Like the old days ...
But in many cases the requirement need not be so time-consuming as full control often demands and it seems to me that the simpler solution (but with a "RAW" file output) is what Capture One has chosen to offer at this point. So the advice I was given about using fewer images made perfect sense and provided a better result.
0 -
Here is one of the non-HDR panos. I had to down size in order to satisfy the 2MB requirement
0 -
Thanks guys, I have filed a request like BeO suggested. Lets wait and see... The hint of SFA does not help in my case but soundet promising. But even with enough connecting points c1 was not able to stitch.
0 -
Both Hugin and PTgui suggest manually adding more Control Points when a pano does not stitch well. The segment of your pano that has the artifact does not appear to have many CPs. The amount of Overlap is not the correct metric; the number of common CPs between the two adjacent images is what determines success or failure of a stitch. If you look at my image (5 images and 40 to 50% overlap) you can see that there is a lot of structure, so having sufficient CPs is not an issue. Here is a PTgui tutorial that discusses CPs: https://www.youtube.com/watch?v=sCAuS4vLCLc. At about 2:00 they start to talk about manually adding CPs.
I did not see where C1 presents an opportunity to manually identify/add CPs.
1 -
Stanley,
C1 does not offer an opportunity to add CPs.
In way I can see that as a good thing as in the distant past, with a number of different pano and blending applications, I spent hours messing around trying to get the "perfect" results using what were less than perfect originals.
Using a simpler approach with other software where things either worked well enough or did not work seemed like a better use of time and acted as a sort of automatic filter of viable and non-viable image sets.
One could see the C1 philosophy as having at least a couple of underpinning ideas in the approach.
Firstly the basis of creating a new "RAW" file from the source RAW files and thus allowing the user to refine the main image processing needs just as with any other RAW file (more or less).
Secondly the expectation that the 21st century concept of collective AI - given enough samples and enough time - offers a basis for understanding image data and how to manipulate them that can be deployed for all users.
The need for understanding the mechanics of CPs and all of the other aspects of matching and merging along with the time to be spent on the task should be eliminated for the majority of viable image sets if the approach succeeds. This would be desirable for many people. If one is in the camp that enjoys the more involved process of personally manipulating a set of images into a merged, matched and blended results produced by hand then it would seem that C1's approach may not be the approach that one would prefer.
Much may come down to whether or not one sees the production of a new, stand-alone, RAW-like image as a big advantage. Or a disadvantage since such a file might be considered to be an enforced extra step prior to the availability of a final output file.
Having become less and less interested in chasing results from Panoramas and HDR merged images in recent years, I am not really fully aware of what the various applications in that market are currently offering. The area of "RAW" file production seems like it may be somewhat unique (or at least rare) but I'm not sure.
I am also not sure how much it really matters to people producing panoramas and merges. If people have becomes used to workflows involving other types of intermediate files and are happy with that approach, this C1 concept (as it currently stands) may not appear to offer them much.
Anyone who obtains contentment from the work of matching pixels and manual blending may well not find anything here to fulfil their desires.
Others may find a simple way to discover new photographic adventures with very little effort or learning required.
-1 -
Maybe the problem is too much overlap. I carried out a test pano yesterday with about 10 images of a harbour scene, with lots of masts, wires etc: deliberately quite a difficult pano. I tried the same in both C1 Lightroom. Using all 10 images, neither was perfect: always a discontinuity somewhere. For both, I then omitted some of intermediate images to reduce overlap in problem areas. It would have been possible to use every other image and still have overlap. It took several goes, but I ended up with a perfect stitch with C1 using 6 or 7 of the images. With Lr, I still had a few broken wires or similar. With a bit of work, I was happier with the C1 result. Everything was entirely with no prior adjustments using either application. I was actually pretty impressed with how C1 performed. Early days, I will try some more soon.
0 -
Absolutely agree. However, there have some instances (using Hugin) where a satisfactory stitch required a little manual assistance. My experience so far is that a high contrast scene with good exposure, C1 does a great job of stitching.
0 -
Overlap is not the correct metric, it is Control Points that count. More overlap will increase the likelihood of finding enough CPs, but if it is a low contrast scene that is devoid of detail, more overlap will not do any good.
0 -
I think the danger with overlap comes when there is potential for more than 2 images to be "present" within the overlap area being assessed. Hence the "less is more" advice in the User guide.
This is most evident with hastily and hand held shots that will be stitched into a multi image width and multi row "panorama". At least it is in my experience.
0 -
You are probably right. But since I shoot my panos on a tripod with a leveler and rotator I don't encounter that problem.
0 -
Here is Paul Reiffer's tutorial on panoramas: https://learn.captureone.com/blog-posts/shooting-for-panoramic-stitching-by-paul-reiffer/
He has several recommendations for using C1,22 for panoramas including 40% to 50% overlap.
Here are several videos on C1,22: https://learn.captureone.com
0
Post is closed for comments.
Comments
13 comments