The large format sacrifice

[torger]

Agreed. Im developing a method as fallow: i have a ICC for each lens/sensor position ( we talk about 3 to 9 stitch ) and i apply to each single raw. Time consuming but this give me a digital large format file of 12000x10000 approx. Which i then resize to 6000x5000.

I'm not 100% sure how C1 does it but I would guess that just like with other LCC-capable programs the LCC application will cause a slight white balance shift, which means that you may have a little bit of white balance variation on the different segments. If that's the case you would notice it in the stitching process though, that the overlapping parts don't match up exactly in white balance.

[Lorenzo Pierucci]

I'm not 100% sure how C1 does it but I would guess that just like with other LCC-capable programs the LCC application will cause a slight white balance shift, which means that you may have a little bit of white balance variation on the different segments. If that's the case you would notice it in the stitching process though, that the overlapping parts don't match up exactly in white balance.

I use photoshop CC for the stitch, and  SEEMS it fix slight expose variations between shoots, maybe ( just maybe ) it do same with the WB ?

[torger]

I use photoshop CC for the stitch, and  SEEMS it fix slight expose variations between shoots, maybe ( just maybe ) it do same with the WB ?

Yes that's not unlikely, stitching software often correct white balance variations.

[Bart_van_der_Wolf]

I'm not 100% sure how C1 does it but I would guess that just like with other LCC-capable programs the LCC application will cause a slight white balance shift, which means that you may have a little bit of white balance variation on the different segments. If that's the case you would notice it in the stitching process though, that the overlapping parts don't match up exactly in white balance.

Hi,

In principle, the brightest part of the LCC image (=approx. optical center) will remain unchanged, thus allowing perfect stitching. But it does indeed depend on the actual implementation if that general procedure is followed.

Cheers,
Bart

[torger]

In principle, the brightest part of the LCC image (=approx. optical center) will remain unchanged, thus allowing perfect stitching. But it does indeed depend on the actual implementation if that general procedure is followed.

When you stitch large format with a comparably small sensor it's not sure that you even have the optical center in the image. And if you use a center filter on the LF lens the optical center may actually be slightly darker than a bit out from it. Another problem is that the MF CCD have color casts in their own, so if you put the optical center in a different position of the sensor you may get a slightly different response. That error should be relatively small though.

I don't know how C1 calculates the reference white from the LCC, it's something one could test, and certainly interesting for those stitching within a large format image circle.

(In Lumariver HDR I've solved it in the way that you can select an "anchor LCC" so the reference white will be exactly the same for all frames.)

[Bart_van_der_Wolf]

When you stitch large format with a comparably small sensor it's not sure that you even have the optical center in the image. And if you use a center filter on the LF lens the optical center may actually be slightly darker than a bit out from it. Another problem is that the MF CCD have color casts in their own, so if you put the optical center in a different position of the sensor you may get a slightly different response. That error should be relatively small though.

Hi,

An LCC is formally nothing more than a normalization of channel response across the image in linear gamma space. It should not influence White balance for that frame. It uses the brightest area of each channel of an image (wherever it is, center/edge/corner/elsewhere) and normalizes that to 1.0 (R=G=B=1.0 or 100%). It then calculates the percentage of relative darkening across the image area for each channel. It then divides each channel by the darkening percentage for that channel (e.g. 90% brightness gets divided by 0.9, therefore boosted by a factor of 1.11), thus creating a uniform response across the image. That normalized data is then used for demosaicing, given a certain user input for White balance.

Because it uses the brightest spot in each blurred channel, it should not change the white balance at that spot. That does mean that an actual White balance difference between shots (light has changed or sensor responds off-neutral at the brightest spot) is maintained. So with extreme shifts which leave the brightest spot off optical center, the white balance will remain shifted.

I don't know how C1 calculates the reference white from the LCC, it's something one could test, and certainly interesting for those stitching within a large format image circle.

(In Lumariver HDR I've solved it in the way that you can select an "anchor LCC" so the reference white will be exactly the same for all frames.)

Nice, especially if applied before demosaicing, although it may increase noise a bit. It will still help if applied after demosaicing (as a kind of color correction), but at slightly reduced color accuracy, although probably still better for uniform color than without it. Post-processing is often about trade-offs.

Cheers,
Bart

[torger]

It should not influence White balance for that frame.

Yes if the LCC algorithm can find the optical center in all images for the stitch and the sensor does not have varying RGB response over the surface this is true. However I'm not so sure that's the case. Looking for the brightest spot can be a too simplistic method, and looking at only one set of pixels too, so you need to average over some surface (to battle noise, hot pixels, etc). This means that the RGB reference if derived from different LCCs won't be exactly the same for all, there will be some slight variation. If the variation is large enough to be a problem or not I'm not sure, it will probably depend on the situation.

I have not yet had time to study large format stitching in detail but it's something I have on my todo list, then I would be able to provide better answer exactly on how big this problem is (or not).

Anyway, I agree with you that in theory LCC does not influence white balance, simple theory. I'm talking about effects due to limitations in precision in the actual implementation.

[Lorenzo Pierucci]

(In Lumariver HDR I've solved it in the way that you can select an "anchor LCC" so the reference white will be exactly the same for all frames.)

Interesting, can u explain more?

[Gigi]

If white balance is set on the shots to a specific fixed temperature, would that help?