View Full Version : Free Chromatic Aberration Correction Software


William Gardner
February 22nd, 2006, 09:00 AM
Hi Everybody,

Various folks have been less than happy with the chromatic aberration (CA) seen on some of the new HD cameras at the ends of the zoom ranges. There is software available which allows the independent radial scaling of the RGB channels of STILL images to partially correct this CA, but I haven't found anything available for video.

So I decided to write my own VirtualDub filter/plugin to do this. And now I'm making it available to all of you, for free of course. :)

You need to be running VirtualDub. The filter plugin, called "CA Correct", can be downloaded at
http://dsp.ucsd.edu/~wgardner/VirtualDub.htm

I've copied the user notes from that site below: be sure to read them and use this correctly!

I hope some folks find this helpful.
Cheers,
Bill
----------------------------
Chromatic Aberration Correction (CA Correct)

What it does: This filter allows you to radially scale (i.e., zoom) the R, G, and B channels of a video stream with independent scale factors. This can be used to correct radial chromatic aberration.

For example, if white objects have a green inner border, then scale the green channel outward with a scale factor >1 to overlay the green channel with the other channels as best as possible. This is similar to what is offered in a number of software packages for still images.

Notes:

1. Since this is radially scaling a frame, it should be used with PROGRESSIVE video. Otherwise, strange blending of fields will occur.

2. To use with INTERLACED video, first apply your best deinterlacer, then use CaCorrect, then reinterlace if necessary.

3. You have to zoom IN on particular channels for correction. Zooming OUT is not permitted, as this would create border effects at the edges of the video.

4. Since the CA is roughly constant only at a fixed zoom position, this software won't be useful for shots where the zoom setting is changing substantially during the shot.

5. And this only does radial correction. Any non-uniform, non-radial aberration won't be able to be corrected by this software.

Tim Brown
February 22nd, 2006, 09:07 AM
That's incredible William. People like you are what makes this a great community!!!

I've not personally purchased one of the new mini-HD cams yet. My company is expecting the arrival of an H1 soon, I'm still waiting on NAB, but am thankful that there are generous people such as yourself willing to give up their blood and sweat for the benefit of all.

Thanks again!!!

Tim

Ben Winter
December 15th, 2006, 12:38 PM
How do you use this when the chroma aberration is on the outside? I've tried scaling up the R and G channels because I'm getting blue fringes on contrasty lines, but I see absolutely no difference...adding the filter makes the progress window when rendering out appear for a split-second and then disappear as if it's done already. Any kind of bug you can fix?

Bob Hart
December 15th, 2006, 06:18 PM
As already advised in point 5 of the initial post, this method may not work as well with footage originated on cameras which do not have the optical centre of radii coincident with the centre of radii generated by the correction software.

Dealing with that would require software becomes capable of self-generating and then mapping from a movable centre, preferably one that generated initially from camera-specific presets like capture software does, then fine tuned itself.

In these instances the CA correction will not be as effectively applied evenly across the whole frame.

Variation between the lens optical centre axis and centre of CCD across camera types and brands exists.

The construction method used in some
lens-in-camera types means there can be variation across a single camera type due to movement or distortion of casework in use and from abuse.

So if results are inconsistent across the entire frame with CA remaining in areas of the frame which mirror each other, don't go after the software writer. There are practical limits to the degree miracles can be wrought.

I am not a industry professional or software writer so if there is anything in my comment which is patently wrong, please slap me down (gently for a soft landing, not "stretched senseless on the floor"**) with an appropriately worded correction.

**excerpt from "How Green is my Valley"

Adriano Apefos
September 16th, 2007, 12:53 AM
Hi, William,

I found your thread about the ca correct in my searches... this solution you developed seems to be amazing. There is a video with the ca problem that I would like to show you. It is a radial ca, so I think your plug in would work. I would like to try it but the download link is broken. If you want to see the video and the thread about this go to:

http://www.dvinfo.net/conf/showthread.php?t=103629

thanks

Adriano

David Delaney
September 16th, 2007, 05:22 PM
The link to William's page no longer seems to work. Anyone have this plugin?

William Gardner
September 17th, 2007, 10:20 PM
Hi Guys,

Sorry for the delay, I had to dig around to find the plugin on a different computer. It is now uploaded to the following link:

http://www.wrgardner.com/CaCorrect.vdf

I just this week saw the post above reporting a problem with it. If other folks have similar problems please let me know and I'll check it out.

Bill

Sun Yang
July 6th, 2008, 11:57 PM
Dear William, thank you very much. Good work. If I try use it for Canon XH-A1, CA on x-axis can be compensated. But XH-A1 has relatively small CA on y-axis. Therefore, if I compensate x-axis CA, y-axis CA will be much worth, then without any compensation. So I want ask you, will be possible produce modified ca correction plug-in with linear CA compensation in x-axis (or independent in x and y axis)?
Thanks