Correcting Sony RX-100 images with Cornerfix.

September 12, 2012  •  1 Comment

This is rather a WTF post... many people have asked about the colour shifts they see on RX-100 files towards  the corners, typically a relative increase in blue (more) and red (a bit less) channels that can give the impression of purple edges. Others have complained of seeing cyan shifts, though I suspect that is due to the way the colour of skies can change naturally towards cyan in certain areas.

To be honest I hadn't given this too much thought - it hadn't bothered me on any of the files I have shot, possibly because I am so used to slight shifts with Leica M9 files and wider lenses and partly because many systems do this 'a bit' but it gets lost in the mix of perception.

Where it really does matter is in shooting wide angled technical lenses on technical or field cameras with medium format digital cameras. In both this case, and the case of the M9 with wides, the phenomenon is caused by the fact that the rear element of the lens is close to the sensor and therefore, rays reaching the peripheries of the sensor are doing so at a sharper angle. Even with specially designed micro lens arrays, this causes both luminance vignetting and colour cast problems. Those of us familiar with medium format digital are accustomed to shooting an 'LCC shot' through a thick sheet of optically correct plastic, for every (if we are diligent) 'regular' shot we make. Later, in post production, we feed the LCC reference shot to Capture One and it subtracts what it finds from the 'real' shot in order to neutralise it.

This is tedious. It is also very necessary with certain lens/sensor combinations to exactly match the focal length, aperture and movements (tilt and shift) of the LCC shot with the 'real' shot, which means that if you want to do it properly, you have to always, always make an LCC file. And when Phase One introduced the IQ180 back, it turned out that many people's favourite tech camera lens, the wonderful Schneider 35XL, was no longer fully correctable and when used with anything other than small movements, fell apart altogether. It became unusable. 

In Leica world, the rather wonderful Sandy McGuffog invented Cornerfix to help. The problem was that though Leica tried to build corrections into their DNG half-baking algorithms, their efforts were doomed to partial success because the camera has no way of knowing the true and exact aperture the lens is set to. So it can access generic lookup tables only whereas what is needed is specific shooting data.

Cornerfix works by taking what is in effect an LCC (Lens Cast Calibration) shot as per Medium Format practice, and doing with it what Capture One does for Phase One files. So you have to shoot a white card (some people use a perspex sheet, like I do, it's more accurate) at the same shooting parameters as the file you wish to correct, and then Cornerfix will in effect subtract the luminance and chroma shifts of that file from your treasured snapshot and voila, your shot is clean.

You do this by shooting your real and LCC files in RAW, opening them in a converter, exporting them as DNG and then using Cornerfix to introduce them to each other in a constructive way. In other words, it's a PITA (not Sandy's fault at all, it's the nature of the beast) and something you'll only want to do when you have to.

Trouble is, you never know you need to until you see colour shifts spoiling your shot - by which time it's often too late to shoot an LCC with exactly the same parameters, if you are a Leica user.

Back to the RX-100. 

Luckily, since the zoom lever tends to stop at discreet and repeatable intervals and since it records aperture to EXIF, you CAN make an LCC later.

Trouble is, it doesn't seem to do anything and I can't work out why.

This shot was taken on a cruise a coupla weeks ago and I thought I might be able to see slight chroma shifts so I decided to have a bash at fixing it:

_DSC0414 So I recreated the focal length as close as I could (very close) and set the same aperture, then shot through an optical grade LCC sheet with +1 exposure compensation (to get the histo about right). This resulted in the following shot:

_DSC0440 Now: this shot has very clear shifts towards the edges. So I applied it to my Chefs in Cornerfix and got this:

NEWDSC0414_CF

But if I look at the original chefs, versus the 'fixed' chefs, in Lightroom's Develop Module, I can position the magnifying glass over the exact same pixel on both files, flick between them and see no difference in RGB values.  Furthermore If I use the LCC shot to 'fix a version of itself', the two are identical.

(clicking on any of the above files will give you a 100% size version at 91% quality JPEG but note that the files look rather flat since I have not applied any tonal corrections, in order not to skew the comparison)

Now, this was not what I expected to find. Possible explanations:

1) I am doing something wrong. Possible but I have been careful to rinse and repeat and to read the instructions

2) There's a bug with the software.

3) The colour shifts are so slight as to fall outside what Cornerfix considers possible to fix (or worth fixing?)

I vote for 3) above. I'd love to hear from anyone who has also tried this experiment but I must say that life suddenly seems too short to dig any deeper: I have yet to see a real-world RX-100 file that is notably compromised by colour shift and so, to all those who have emailed me raising the concern and asking me for opinions and cures I say, go shoot! But if you have a file that really is badly affected, try the method I have suggested above: colour shifts vary with aperture and focal length and some combinations might give enough trouble to be worth CornerFixing... for the rest, I suspect that either Lightroom, or Sony in-camera, is already doing most of the lifting...

 

 


Comments

canvas prts(non-registered)
totaly beautiful, wow, whatr fantastic photos.
No comments posted.
Loading...
Subscribe
RSS
Archive
January February March April May June July (24) August (11) September (8) October (4) November (5) December (4)
January February March (2) April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December