Flat Stitching
Please support flat stitching of frames with individually offset LCC and lens corrections. This is for files captured with a tech camera and rear shift.
-
...that is a very good request, but I think we might need to be more specific as to how it is implemented to be applicable to as many use case scenarios as possible.
In your case when working on a technical camera doing a shift of the image plane there is no way of telling what portion of the image circle you are in so you would have to create LCC data for every imaginable position of your shift which can be a hassle and might not yield the results you are hoping for.
Maybe a more practical approach would be to be able to create an LCC from a finished panorama stitch (or any image file for that matter) and apply it to another stitched panorama with similar dimensions.
Currently only LCCs from the same camera/lens combination are possible. I can't even create an LCC from one of my A7RIV and apply it to an image taken with another A7RIV, it has to be the same. But if it were possible to make an LCC from any image and apply it to any other image with roughly similar dimensions you could create LCC's for a wide array of situations including your case.So to sum it up:
Have the possibility to create LCCs from any image file (without dust removal information and wide angle distortion data obviously) and be able to apply it to any other image file, even if they are not from the same camera/lens combination.0 -
I don’t understand your idea of a universal LCC file and how it would apply here, because the necessary corrections vary with lens, shift amount, focus distance, aperture, and digital sensor. It’s part of the photographer’s job to capture these files, note the shift amount, and enter the appropriate info in the raw converter (C1).
Without cameras that record shift amount, focus distance, and aperture in metadata, it’s going to be a somewhat manual task. (Yes, I’m aware of the IQ4, XT, and X-shutter. I’m talking about the majority of Arca/Alpa/Cambo/Sinar/Linhof/etc tech cameras that exist with mechanical-only connections.)
This is not a new idea, it’s been standard practice with tech cameras and digital backs for a long time. Here’s an example workflow for a two-shot flat stitch, 30mm apart:
- Set up camera, focus, and stop down lens.
- Shift back right 15mm*.
- Make a normal exposure.
- Place white plastic sheet in front of lens, make LCC capture.
- Shift back left 30mm (so it’s +15mm from 0).
- Repeat steps 3 and 4 in this position.
- Load all 4 exposures into Capture One
- Assign appropriate X shift values in the lens correction tab: here it will be +15mm, -15mm*. Do this for all 4 shots, including the white LCC shots.
- Select the two LCC shots, choose “Create LCC” with “Include Wide Angle Lens Correction Data” checked.
- Select the pair of standard shot and LCC shot for one shift position.
- Choose “Apply LCC”
- Repeat steps 10 and 11 for the next shift position.
- Select the two standard shots, now with LCC correction applied, and choose “Edit with Photoshop”.
- In Photoshop, stack the two images into one file, and Auto-Align Layers, Reposition.
- Save as TIFF or PSD file.
Request: Perform steps 13-15 within Capture One, resulting in a DNG file of the combined frames.
*In C1, think of XY shift values as lens movement. From the perspective of the photographer behind the camera, left lens shift would be +X, right lens shift -X, front rise would be +Y, front fall -Y. These +/- directions will be opposite for back movements.
0 -
Hi Alan
I understood what you wanted, but I don't think you did understand what I suggested, most likely because I wasn't clear enough in my explanation.
I wasn't asking for a universal LCC profile, but simply for the ability to:
1. create an LCC (as in "lens cast correction") profile from any image file regardless of format (raw, tiff, jpeg, etc), pixel dimensions or aspect ratio since all we need is information about light falloff and color cast.
2, apply that LCC profile to any other image, regardless of format (raw, tiff, jpeg, etc), pixel dimensions or aspect ratio since all we apply is light falloff and color cast. Currently you can only apply a created LCC profile to an image that comes from the same camera/lens combo (though DNG as camera and "no lens data" would work as identical camera lens combo, as illustrated below)So if you want to do your 12 step procedure of creating an LCC profile for every one of your individual shift positions you still can and you could apply each of the created LCC profiles (-15, 0, +15) to the corresponding shots of your panorama individually before your position 13. and then do a distortion correction in photoshop (if that is your tool to do the stitch) once the stitch is done.
But if you didn't want to do the multi step procedure that you are proposing and that incidentally would require a whole new and rather complex tool within capture one, you could create an LCC profile from a completed panorama stitch of a white uniformly lit surface and apply it to your finished panorama (I use a repro table for that and it works great)
I don't know if you noticed, but C1 averages the image file used to create an LCC profile so no details are retained, just the light falloff. I use the LCC function quite regularly when reproducing art works for artists in their studio. I usually pick a roughly uniform wall to hang the art works on, put up some lights, don't even bother to be too precise about it, because once I shot an image of that wall I will use it as LCC data. Unlike the "Flat Field" function in LR, C1 is not including any textures or blemishes found in the reference image to make the LCC profile (unless you check dust removal data). So if you were to create a stitched panorama of a greyish surface with a little bit of texture on it, all C1 would retain for the LCC is the actual light falloff and color cast, not the texture. So you can then use this to create a stitched image of your grey surface (because there is actual texture on it to make an automatic stitch and create an LCC of a finished panorama, without having to enter any lens shift information.
Currently I cannot create an LCC profile from a DNG stitched with C1's new panorama function, though weirdly I can create an LCC profile from the same panorama stitched in Lightroom, so all I am asking is to be able to do that within C1. I just use a lot of words.
I attached a few screen shots to illustrate my process.
1. individual shots of a uniformish surface with texture on top and corresponding panorama on the bottom.
(I intentionally just lit one light of the repro table to simulate a light falloff)
2. Stitched panoramas imported from Lightroom (it is important to check "Fill Edges" in Lightroom when doing the stitch, otherwise C1 will do strange things to the edges when creating the LCC)
3. created and applied LCC profile to grey surface and panorama
4. and zoomed in you'll notice that none of the pencil line on the grey surface are retained when the LCC is applied to the panorama, just the light falloff.
...and just in case this wasn't clear: I used the grey image on the left to create the LCC, but any roughly uniform surface would do. The whole thing you are doing with the white plastic in front of your lens is how one used to do it, but it is no longer necessary as once again no textures are retained when creating an LCC in C1.
0 -
OK, I think I'm catching on to what you're asking for - applying LCC post-stitch? These things take longer to type than to actually do!
With your example, are you rotating the whole camera to get the individual shots, or using lens/back shift?
0 -
...for the example I used a Sony A7RIV on a Cambo Actus with a 80mm Schneider lens and a 20mm shift between each of the three shots, resulting in a panorama of 36mm x 64mm or roughly 9500x17000 pixels.
The advantage of the LCC post-stitch is that it is easier (and therefore more likely) to be implemented as a function in C1 and more flexible in use as it can also be used for light falloff and color shift data that isn't actually caused by the lens but by uneven and differently colored lights as in the art work reproduction example I mentioned. In actual use I also found it to be much faster and more accurate than doing LCC corrections on the individual shots of a panorama made on a technical camera.
0 -
Sounds like a good option.
For distant scenes, the LCC frames have to be created with a white filter on the lens. Since there's no textural info as in you copy stage example, the individual XY offsets would have to be specified. For applications like architecture photography, it's common that none of the individual shots are centered, again requiring XY offset info.
Although we've only been talking about falloff and color cast, CA and distortion corrections are also influenced by XY shifts.
0 -
...all i can say is that you should try it, before assuming that it doesn't work because an LCC profile created from a white surface taken at 50cm away is evidently different from an image taken of a landscape or a building at several hundred yards. But most of the time when looking into LCC profiles for the kind of application you are talking about, we are looking at short to very short throw lenses often of symmetrical design with very strong color cast towards the edges due to bleeding of light of the bayer filter on top of the individual sensors into adjacent sensors and while there is a difference in the color cast at 50cm or 50m the difference in practice is so minor that you will be hard pressed to find any trace of it once you apply it to your image (all build in lens profiles work with the same level of latitude). As for the different use case scenarios you are describing with architectural shots not being centered, nothing prevents you from creating an LCC after the shot by simply replicating its offset on a repro table the way I described it.
I understand your doubts, I had them too, but you'd be surprised how well this works once you try it.My gut feeling is just that given the direction C1 is heading a complex tool like the one you are describing would take some resources and commitment to develop for something that has in reality a very limited, niche use. On the other hand being able to create and apply an LCC profile from any file to any file would get you 95% of the desired functionality at about 5% of the effort to develop it, since it is in fact already there (as illustrated with the pano imports from lightroom), just not yet made available for panos created inside C1. Once there is a functioning workaround I have no issue to wait the years it might take before the tool you ask for gets on top of their priority list.
0
Post is closed for comments.
Comments
7 comments