CornerFix 1.2.0.0 available for download

[sandymc]

Advertisement (gone after registration)

CornerFix 1.2.0.0 is now available. This version significantly improves image quality when correcting M9 images, and fixes the compatibility issues with Capture One V4.8.3.

 

Note that this version has considerable internal changes; treat it as a beta release.

 

In detail:

 

1. There is a workaround for the Capture One 4.8.3 bug that resulted in images being read in with a color cast or looking overexposed

 

2. Black level handling has been optimized to deliver higher image quality on conversions, especially at high ISO

 

3. CornerFix now has an additional option for profile generation - to do either "bidirectional" or "unidirectional" correction. A quirk of the M9 that became apparent in testing is that it is very easy to get overcorrected "pink corners" in images. Whereas previously CornerFix would allow slight overcorrection in parts of an image if, on average, it resulted in a better overall result, the new version now by default avoids any overcorrection.

 

The sample images below show the improvement for an uncoded f/1 Nocti, first the original image, then corrected with the previous version of CornerFix, then the new version. (Original image courtesy of Bob Parsons)

 

Note however that with the new version, if you want to correct for over-correction by the camera, e.g., because you miscoded the lens, or had an M8 set to UV/IR without a filter, you MUST enable "Bidirectional Correction"

 

Thanks to all those that helped testing and/or used up valuable time with their expensive new toy pointing it at blank walls

 

You can download from: CornerFix | Get CornerFix at SourceForge.net

 

BTW, if you get an obnoxious add suggesting that you provide any personal information to download a useless "white paper" or similar, just ignore it.

 

Enjoy,

 

Sandy

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[adan]

Thanks! - and a background question, since you seem to know your "corners". I take it the red tint in the corners of M9 originals with some lenses is likely an overcorrection by Leica's firmware? Since I'd expect an internal sensor IR filter to produce green corners (as with the M8/external filters) if left to its own devices.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[sandymc]

Andy,

 

Yes, that's right - there seems to be something about the M9 that makes it prone to overcorrection in the corners. I've seen quite a few images with red on the sides and corners. Also Sean's implied a few times that, at least with the CV12 and 15, the M9 is a bit of a handful - this may be what he was referring to.

 

Perhaps the microlens layout or something. But it starting to feel like the M9 may be less forgiving that the M8 in regards of correction for cyan drift. But in many ways, we're still in the early days of understanding how to get the best out of the camera. No doubt this is one of the issues that will become clearer over time.

 

Regards,

 

Sandy

[docolmo]

It seems even on the M8, the camera is overcorrecting the IR effect on the 18mm super elmar f3.8.

[sandymc]

It seems even on the M8, the camera is overcorrecting the IR effect on the 18mm super elmar f3.8.

 

Yes, several people seem to be having that problem. CornerFix should be able to correct for that however - the instructions are in the PDF file supplied along with CornerFix.

 

Sandy

[adan]

I noticed Stephan Daniel refered to the issue of overcorrection in the LL video - one reason why Leica does not correct for "natural" vignetting - i.e. the type that is lens- rather than sensor-based, and would occur on film as well.

 

I suspect there will be tweaks to the M9 firmware lookup tables as Leica gets more real-world feedback and experience with the M9. I don't know if that will make your life easier - or more exciting, Sandy!

[sandymc]

Advertisement (gone after registration)

I don't know if that will make your life easier - or more exciting, Sandy!

 

Always the question!!!!!

 

However, I'm starting to get the feeling that it may actually be the "almost there but not quite there IR correction" on the M9 that's going to cause the most complications; you find yourself in a bit of a catch-22. No external filter, and you might just get IR contamination that will keep you busy in post for a (very) long time. Put an external filter on, and you'll be busy in post as well, although not as long. Tricky.

 

Sandy

[sean_reid]

I noticed Stephan Daniel refered to the issue of overcorrection in the LL video - one reason why Leica does not correct for "natural" vignetting - i.e. the type that is lens- rather than sensor-based, and would occur on film as well.

 

I suspect there will be tweaks to the M9 firmware lookup tables as Leica gets more real-world feedback and experience with the M9. I don't know if that will make your life easier - or more exciting, Sandy!

 

Stefan was speaking about luminance vignetting correction. The slight over-correction for cyan drift is present even when some wide lenses are used on the M8. Its also possible that these corrections work better under some lighting conditions than others.

 

When the camera is set to "auto" in the lens detection menu, my tests show some over-correction of cyan drift in files made with the WATE set to 21 mm. But of course the first solution to try there would be to manually set the camera to "WATE at 21 mm" so as to back off the correction.

 

I'll stress again that it is important (in my view) that the M9 *not* automatically over-ride a manual lens selection when a coded lens is mounted. By using the manual lens menu we can dial in corrections, for a given mounted lens, that are stronger or weaker. So, with the Leica 18, for example, we could try "cross coding" (to make up a new term) it as a 21 mm lens and that may fix any overcorrection we're seeing when the lens is simply auto-detected.

 

As a rule, cross coding to a longer focal length will decrease cyan drift correction whereas cross coding to a shorter focal length will increase cyan drift correction.

 

Lastly, its worth keeping in mind that new firmware versions may well included revised corrections for the various lenses.

 

Cheers,

 

Sean

[ho_co]

... So, with the Leica 18, for example, we could try "cross coding" (to make up a new term) ...

Gee, Sean! I had forgotten that the term 'cyan drift' could disappear from the lexicon with the elimination of the need for UV/IR-Cut filter for every lens.

 

I'm glad you've got a new one for us--and this one probably won't require your constant explanatory intervention!

 

 

 

You gotta feel sorry for all those filter companies who aren't going to be getting rich by selling quantities of these weird designs any more!

[reven]

I really wish Phase would give us access to LCC files with DNGs. However, that probably won't happen. Am I correct that CornerFix works similar to Capture One when creating LCC files ? Or is there a different idea behind it ?

[sandymc]

I really wish Phase would give us access to LCC files with DNGs. However, that probably won't happen. Am I correct that CornerFix works similar to Capture One when creating LCC files ? Or is there a different idea behind it ?

 

I'm afraid I have absolutely no clue at all how Phase One create LCC files. Or any idea what's in them. Sorry!

 

Sandy

[scott kirkpatrick]

Stefan Daniel, in the video interview, made a comment that the built-in vignetting corrections in the M9 do not depend on aperture. Does that indicate that the M9 now only corrects for microlens effects and red vignetting (cyan drift) both of which should be aperture-independent. Lens vignetting has a baseline correction and then a stronger correction usually for the widest 2 stops where the lens tube blocks some of the light that should have reached the corners. Is that extra part left uncorrected?

 

Finally, red-cyan effects due to IR are more pronounced in incandescent illumination than in daylight, and hard to predict in flourescent or mixed light. Does Leica anywhere say what color temperature their corrections are optimized for?

 

scott

[Mauribix]

It seems even on the M8, the camera is overcorrecting the IR effect on the 18mm super elmar f3.8.

 

I noticed the same behaviour with the M8 depending on the aperture set...

[Mauribix]

Actually the worst in my lens set seems to be applied to the Elmarit 21ASPH...

 

I thanks Sandy, this new release will surely come in hand.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[tashley]

I really wish Phase would give us access to LCC files with DNGs. However, that probably won't happen. Am I correct that CornerFix works similar to Capture One when creating LCC files ? Or is there a different idea behind it ?

 

I have asked them to consider doing this in their next version. It really works well on my Phase P45+ when used on a field camera with wide Schneider glass. It would be perfect for the M9 / CV15 combo...

[tashley]

I'm afraid I have absolutely no clue at all how Phase One create LCC files. Or any idea what's in them. Sorry!

 

Sandy

 

Technical cameras when used with wide lenses create shifts like we see on the digital Ms. This is exacerbated when you use movements of the lens relative to the sensor such as shifts designed to allow landscape and architectural photographers to use more of the lens's image circle. To overcome it, you shoot an LCC reference shot for every normal frame (unless you're lazy like me and keep a library of 'frequent flyer' files). This is done using a thick, optically and colour neutral perspex diffusion sheet (and boy are they thick. And expensive.)

 

You then tell C1 to Analyse this file and apply the resulting LCC profile to all shots taken with that lens, aperture and degree of shift.

 

Phew. But it works!

 

Tim

Edited October 1, 2009 by tashley

[reven]

It works and it it works amazingly. If it was possible to use it with DNGs we could actually use nearly every lens with any color cast. Just need to shoot a reference a few times. I would should it at a few apertures and light conditions. I only have the feeling phase is not caring to much after the Leica split.

[sandymc]

Technical cameras when used with wide lenses create shifts like we see on the digital Ms. This is exacerbated when you use movements of the lens relative to the sensor such as shifts designed to allow landscape and architectural photographers to use more of the lens's image circle. To overcome it, you shoot an LCC reference shot for every normal frame (unless you're lazy like me and keep a library of 'frequent flyer' files). This is done using a thick, optically and colour neutral perspex diffusion sheet (and boy are they thick. And expensive.)

 

That's very similar to what CornerFix does, with the exception that CornerFix is optimized for corrections that are symmetrical around the center of the image. So CornerFix would not be able to correct for lens movements, etc.

 

Sandy

[tashley]

That's very similar to what CornerFix does, with the exception that CornerFix is optimized for corrections that are symmetrical around the center of the image. So CornerFix would not be able to correct for lens movements, etc.

 

Sandy

 

That's interesting Sandy. Could you not make it take asymmetricality into account just by showing it a reference image? I ask because as you'll have seen, every lens, even different sample of the same model, has a different signature pattern of shift...

 

Best

 

Tim

[Advertisement]

Hallo,

Look here: Leica M

[sandymc]

That's interesting Sandy. Could you not make it take asymmetricality into account just by showing it a reference image? I ask because as you'll have seen, every lens, even different sample of the same model, has a different signature pattern of shift...

 

Tim,

 

I could, in principle, but it's a trade-off. I did in fact at one point build a version of CornerFix that did correct for asymmetry, but it was never released, as on average the results weren't as good as for the symmetrical only version. The issue is not actually the correcting part of the process, which is easy, but the profile creation part of the process. The advantage of CornerFix effectively ignoring anything except radial variations is that it makes CornerFix a lot less sensitive to uneven lighting than would otherwise be the case. Given that CornerFix is mostly used by people that don't have access to high-end studio equipment, this allows people to generate pretty good profiles fairly easily - e.g., just taking a photograph of a blank wall. Versus the thick expensive acetate(!)

 

Regards,

 

Sandy