I've made some tests of the Hasselblad Phocus flatfield correction (LCC correction) algorithm. I have a H4D-50 to test with, and SK35 as the widest angle.
For tech camera users flatfield correction is a very important feature, as for all semi-wide to wide lenses will lead to a cast on the sensor, especially if shifted. Phocus has support for this, they call it "Scene calibration".
A basic implementation is easy to do, you just divide every pixel in the original image with the LCC image, but that will increase noise and thus require you to have a very well-exposed LCC image. What converters then generally do is to smooth the LCC image which takes away noise, but then you get problems with cancelling out dust spots and any specific sensor artifacts like tiling or microlens ripple.
I'm quite sure that Phase One's Capture One and Lumariver HDR are leading concering advanced LCC corrections and handles all those challenges (I wrote the algorithm in the latter, and of course that's the best
). But where does Phocus stand?
Phocus has a simpler algorithm with just smoothing the LCC image so they can't remove dust spots, but is robust concering noise in the LCC image, and as it's simple it's also fast. Although I have no Dalsa-based H4D-60 to test with I'm sure it can't remove specific sensor artifacts (tiling, ripple), so you'd want to use a sensor without those issues.
I've heard some say that Phocus LCC over-corrects vignetting (ie make corners too bright), but I have seen no such issue. However, all lenses vignette at least little so it's how we're used to look at images so cancelling out it fully may not be the best thing to do. I usually add back about 0.3 stop artistic vignetting after LCC correction. Unfortunately there's no artistic vignette tool available in Phocus, so you will have to use another tool for that.
An other advanced aspect of flatfield correction is how highlights are handled. When flatfield is applied samples are scaled. If the algorithm is basic this means that if you have something bright in a dark region (to the side in the image circle) it may become clipped after correction, and indeed this can happen with Phocus. You should therefore certainly use center filters for your wide angle to minimize vignetting and thus this problem. As the H4D-50 handles shifting very well I would only expect this to be an issue with the SK28 (where you get some pixel vignetting despite center filter), if you use center filters. It is a problem with the SK35 (tested) and probably SK47 too if you don't use a center filter. For the Dalsa-based H5D-60 or the Sony CMOS-based H5D-50c I would expect this to be more of an issue, even with center filters as they have more vignetting and more colorful casts.
There's also the opposite case when you have something clipped in a bright region (close to the center of the lens) it may then become unclipped and look dull. In a worst case they could be discoloured but Phocus at least makes sure that it's neutral. I've noted that there may be some slight discolourings and artifacts in the transition into the clipped area after correction. Again, this issue is battled by using center filters to make the need of correction as small as possible.
Finally, what happens if you shift all the way to hard vignetting, that is if you shift so far that a side or corner of the frame gets outside the image circle and cut to black? Capture One can handle this, but creates artifacts near the edge, Lumariver HDR does this well, but Phocus cannot handle it at all, it refuses to create a calibration frame if the LCC shot has hard vignetting. For most this won't be a problem as it's not a normal use case with such extreme shifts, but if you work with maximized shifted panoramas and crop in post it can possibly cause a bit of irritation now and then.
When the algorithm corrects color cast it needs to get a reference for white. An advanced algorithm may figure out where the center of the lens is (even if it's shifted, and even if it has center filter applied) and take the reference from there as the color cast is least there, while simple algorithms generally just use an average over the whole image or picks the reference from the center. The average works well for mild color casts, and with the H4D-50 you can't get anything than mild casts, but with say a H5D-50c which can have very colorful casts it's likely that you can get visible white balance shifts. With a white balance picker or manual adjustments you can bring it back to normal though. The only problem is that the Phocus color model that uses white balance to figure out which illuminant it's under can be a bit disturbed, but I would think it's not a big issue.
I haven't been able to test how Phocus derives its white balance reference, but as the algorithm is overall basic and not seemingly designed for shifted lenses I'm quite sure that either it just picks the reference from the center of the image (assuming no shifting has been applied), or it makes an average over the whole image. With the H4D-50 you will not notice any issues as the sensor has mild color casts, but you might with other sensors.
The user interface is designed for the case when you have a library of pre-made flatfield corrections, rather than shooting one per image when out in the field (which is the typical case for tech camera users). It works in the latter case too, but your list of "scene calibrations" will grow.
Vignetting correction and color cast correction can be applied separately.
Overall I think the flatfield correction algorithm is a bit disappointing for a technical camera user. With a sensor with such low color cast as the H4D-50 has you won't have any real technical issues as long as you use center filters, but I still find it disappointing that it cannot clean up dust spots, and I'd really like to be able to add in artistic vignetting after correction. The flatfield correction capability is certainly not a show stopper, but a thing to consider for users going the Hasselblad way. If you're used to Capture One's capabilities you may be a bit disappointed.
When it comes to CFV-50c/H5D-50c I have no files to test with. I expect results to be generally good, as there's no artifacts like tiling or ripple with it, but it has really colorful casts and some issues around highlight handling and white balance reference may occur. With the H4D-60 I'd expect that using SK35 and SK28 which some actually do with Capture One, is a show stopper due to no correction of tiling and ripple, but it should work with Rodenstock wides if not shifted too far and not too contrasty post-processing (as far as I know Sean Conboy is using a 60 with Rodenstock wides).
