problem Correcting Chromatic Aberations (CA)? [DSLR Related]

[Stereodude]

I bought a Tokina 11-16mm f/2.8 (Canon) after seeing the Photozone.de review which praised it's high uniform sharpness compared to the other similar lenses. The downfall of the lens is worse CA than other similar lenses. However, everyone always glosses over CA with the ol comment of, "Well you can easily correct CA with most RAW converters." I shoot RAW almost exclusively, so I figured "no problem"...

So I got the lens from B&H yesterday and I shot a few test pictures this morning, and sure enough it has CA. My lens is different from the lens Photozone.de tested though. Theirs had uniform CA across the entire image. Mine has no CA in the center and then it gets progressively worse towards the edges.

Canon Digital Photo Professional has CA correction, but it assumes the CA is uniform across the image and applies a uniform correction. If I dial in the right setting to correct for the CA on the sides of the image, it adds CA to the center where there originally was none.

Adobe's RAW converter (in CS3) behaves similarly again applying a uniform correction across the image. By enabling the fringing control to either all or highlights I can sort of make it work by dialing in a mid way setting that reduces it on the edges and adds some to the center of the image and then the fringing control fixes what's left after the slider adjustment.

Canon DPP is a non starter. ACR 4.x sorta works, but seems less than ideal in my mind. So, my question is how do I correctly compensate for the type of CA my lens has? Are there other programs that can correct for a progressively worsening CA that I should look into, or do I just not know how to use the two RAW converters I have?

If I can't find a good solution I will return the lens. And right now I don't think I have a good solution.

 

[Stereodude]

Here's a JPEG straight from the camera that shows the issue in the silhouettes of the trees. FWIW, the CA seemed more pronounced in the RAW file.

 

[udaman]

Here's a JPEG straight from the camera that shows the issue in the silhouettes of the trees. FWIW, the CA seemed more pronounced in the RAW file.

did U try photo forums? Think that would be better???

 

[Stereodude]

did U try photo forums? Think that would be better???

I asked beyond just here. However, there are a lot of DSLR users here, so I figured I wasn't pushing the limits of the knowledge available here.

 

[LunarMist]

Can you put an example .cr2 file online somewhere? I can try a few things.

 

[Stereodude]

Here's one.

I must have been smoking something funny this morning. Tonight I was able to get much better results with ACR 4.6 in terms of CA removal. :dunno:

 

[LunarMist]

Ugh. It's not easily and uniformly correctable, but can be improved to a reasonable degree. Unlike typical Canon wideangle red-channel offset and spreading, this one has a blue-channel fringe and green-channel offset. Some other apps reduce CA by decolorizing certain areas rather than changing magnification, but they occasionally cause odd artifacts.

We had a Tokina 12-24 for N/AF a few years ago and found it to be problematic as well. Is there some reason not to use the Canon 10-22?

 

[Stereodude]

I forgot to mention that I don't like Adobe Camera Raw because the output has posterization in the shadow details that Canon DPP doesn't.

The Photozone.de review of the Canon 10-22 didn't exactly excite me. The Tokina has better sharpness and supposedly the CA was easily fixable. The Canon also has a fair bit of CA.

So it looks like I can either have CA, or have posterization in the shadows (and whatever else ACR does poorly). I guess I should try out a few other RAW converters. :dunno:

 

[LunarMist]

Convert to neutral 16-bit TIFF in DPP and then do everything else in PS.

 

[Stereodude]

Convert to neutral 16-bit TIFF in DPP and then do everything else in PS.

I'm not sure I follow you. What's a neutral 16-bit TIFF?

 

[LunarMist]

Picture Style: Neutral

 

[Stereodude]

Picture Style: Neutral

How does that help me get rid of the CA though?

Normally I convert to 16-bit TIFF with DPP and do the rest in PS.

 

[LunarMist]

It doesn't. I thought you were on to other issues now. Canon wideangle zooms are not the greatest.

 

[Stereodude]

No, I'm still trying to find a workable solution for nuking the CA and getting a decent RAW conversion. If I can't find one the lens goes back. Right now I'm leaning that way.

 

[ddrueding]

I haven't had that many issues with the 10-22. It isn't exceptional, but challenges me

There is no ca correction in this one.

 

[Stereodude]

Well, apparently Canon DPP CA correction controls are just counter intuitive and not well explained. It is possible to basically dial out all the CA this Tokina 11-16mm lens has with it. You have to set the main slider to 0 and use the lower two. You would think that having the main slider set to 0 means no correction but that just isn't the case.

 

[Handruin]

I haven't had that many issues with the 10-22. It isn't exceptional, but challenges me

There is no ca correction in this one.

But there is slight CA in that picture. Look at the right edge of the left-most clock tower. There is a slight pink/purple fringe on the edge.

 

[Stereodude]

Here are the setting that dialed it out for me in Canon's DPP.

If Imageshack decides to start accepting uploads again any time soon I'll post before and after comparisons.

 

[e_dawg]

Yep, Tokina has a well-earned reputation for CA and flare problems. IMHO, it's not practical to solve "non-linear" optical flaws like that with software. It will drive you up the wall. I would suggest that the problem and solution both reside in hardware: either get a better lens (the 10-22 LM suggested is probably your best bet, not that you really have much choice) or use a different system for WA photography (that is, if WA photography is important enough to you to make it worthwhile).

 

[LunarMist]

Some lenses are easier to correct across the frame than others. Wide zooms are usually difficult compared to prime teles. Of course LoCA is not really correctable.