A Question about correcting Chromatic aberration

pathfinder · March 30, 2010

I got asked by Liquidity the following question:

"I was thinking. Would it be possible to remove CA via the "Impossible Retouch"? In other words, we paint a color over whatever fringe we want to eliminate and then use the "blend if" sliders to change the colors where we painted (to remove the CA)?

If so, would we want to stay in LAB? Additionally, which sliders would we want to fool around with to get rid of the aberrations (i.e., the lightness, A, or B channels)?"

While it should be possible to use blend if sliders to limit where we want the correction for the color fringing to take effect, chromatic aberration is not limited to a small limited area of an image, but extends throughout the peripheral portions of an image, and may affect the vertical image elements differently than the horizontal elements, or the oblique elements.

Adobe Camera Raw already offers a lovely way to correct chromatic aberration that works rather well, and quickly, as well. I wrote my technique for correcting chromatic aberration in ACR here - http://www.dgrin.com/showthread.php?t=159524

The same technique is available in Lightroom, that does not offer access to LAB channels.

Wil Davis · March 30, 2010

Thanks for the heads-up!

That is certainly a useful trick! I try to avoid CA like the plague, and when it does occur I really have no way of dealing with it (apart from saying it was intentional…

)

You really do come up with some cool stuff!

- Wil

pathfinder · March 30, 2010

Wil Davis wrote:

Thanks for the heads-up!

That is certainly a useful trick! I try to avoid CA like the plague, and when it does occur I really have no way of dealing with it (apart from saying it was intentional… )

You really do come up with some cool stuff!

- Wil

Will if you know how to avoid chromatic aberration I am all ears, but I have a number of Canon L lenses, and they all can display some degree of CA in some images if inspected at 300% carefully in the corners of the images. Most of the CA is easily removed in ACR as I described.

I just don't know of a way to "avoid it"

BinaryFx · March 30, 2010

pathfinder wrote:

I got asked by Liquidity the following question:

"I was thinking. Would it be possible to remove CA via the "Impossible Retouch"? In other words, we paint a color over whatever fringe we want to eliminate and then use the "blend if" sliders to change the colors where we painted (to remove the CA)?

If so, would we want to stay in LAB? Additionally, which sliders would we want to fool around with to get rid of the aberrations (i.e., the lightness, A, or B channels)?"

I missed the original post that this topic is referring to. Link?

This "impossible retouch" sounds like it is referring to the Dan Margulis Lab mode retouching technique.

Lab mode can do some things that take much longer or or harder to do in RGB. When it comes to the blend if sliders in the above question - the advantage of Lab mode is that it separates colour from tone via the channel structure - while in RGB one separates colour from tone with blending modes, which are similar in concept but different in practice.

The Lightness channel in Lab and the default RGB luminosity blending (blend if gray) are very similar to each other, so no real benefit one over the other.

Blend if sliders in RGB using the separate R, G or B channels do not work the same as the A or B channels in Lab mode - there is a great advantage to perfroming some blending operations in Lab due to the nature of the Lab channel structure.

The A channel in Lab mode contains the magenta/green opposing colours.

The B channel in Lab mode contains the blue/yellow opposing colours.

So, in theory, yes, stay in Lab mode and use the channel structure of Lab mode to your blend if advantage. In practice, I have not tested CA removal to see if Lab would be a major improvement over traditional RGB edits.

Without testing, my money is on Lab mode and blend if.

For those that are asking what Photoshop layer "blend if" is all about - an introduction to the topic can be found at my site:

http://members.ozemail.com.au/~binaryfx/blendif.html

Stephen Marsh

http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/

pathfinder · March 30, 2010

Stephen,

the original question I received via PM from Liquidity.

Are you aware of a technique to correct chromatic aberration in LAB?

Is that preferable to correcting as metadata in Adobe Camera Raw?

BinaryFx · March 30, 2010

pathfinder wrote:

Stephen,

the original question I received via PM from Liquidity.

Are you aware of a technique to correct chromatic aberration in LAB?

Is that preferable to correcting as metadata in Adobe Camera Raw?

Pathfinder, your tutorial should be moved into the Grad School!

I think that the CA tool in ACR is in the right place (rather than using the similar option in the lens distortion filter in Photoshop). And yes, metadata/parametric editing is great. I think that this is the ideal and best place to perform this task for the majority of images and end users (on raw data if available). Who wants to do this manually, unless there is a good reason to?

The original question was basically asking if it was possible to manually paint or clone in color blend mode to remove the fringe - the answer is of course, yes (one could also desaturate). Further, it was being asked if the special channel structure of Lab mode would lend itself to using the blend if sliders to better isolate the correction. I answered, in theory - yes, as the colour channel structure of Lab mode can make blend if way more powerful than the luminosity (gray) or separate colour channel structure of RGB (this being image dependent, such as a yellow object on a blue background - or any warm on cold or cold on warm hue).

Stephen Marsh

http://members.ozemail.com.au/~binaryfx/
http://prepression.blogspot.com/

pathfinder · March 31, 2010

Thank you for your kind comments about my CA thread.

I agree with the use of LAB and blend if sliders. They can be very useful as a means of local image correction which does not require a selection. I just am not familiar with a technique along that line to correct chromatic aberration. I can see how that might be used for small local areas of residual CA not corrected in ACR, but I find the correction in ACR adequate for most of my lenses most of the time.

DXO Optics does correct for chromatic aberration based on a specific lens profile, I believe. I do not own or use DXO Pro, but my understanding is that the software must have a "profile" for the specific lens to allow correction for CA as well as barrel or pin cushion distortion, etc.

LiquidAir · March 31, 2010

The best way to correct CA is separate geometric correction on each color channel (R, G,

, because the root cause of the problem is that the lens behaves differently at different wavelengths. The ACR correction, I believe, is based on the presumption that the focal length of the lens varies slightly across the spectrum and the slider lets you specify how much. DxO does a full geometric correction for the lens and can do better than ACR but the difference is only noticable on some lenses.

pathfinder · March 31, 2010

Ken, that is kind of what I have been dancing around discussing, that chromatic aberration is due to the different indices of refraction for different wavelengths of light, and hence, as you said, the spread of magenta/ green or blue/yellow varies from the center of the image, getting worse nearer the corners of the image. It is may be visible around large objects, or surround every leaf in a tree image, depending on the lens, lighting and focal distance. As I said in my tute, ACR MAY not remove all of the CA, but will usually make it substantially better. DXO can do a better job for a specific lens, if a profile is available for it.

Since DXO requires profiles for the lens correction, I am not certain if they even offer them for non-OEM lenses, like Tamron or Sigma.

A Question about correcting Chromatic aberration

Comments