I ask because that approach (auto-alignment of R G and B channels to reduce lateral CA, with no knowledge of the lens) is exactly what in-camera JPEG engine does on recent Nikon bodies; and that correction is also enabled by default in the Nikon CaptureNX2 raw converter.rjlittlefield wrote:I don't know if it provides CA suppression for lenses that it knows about, where the CA parameters are known beforehand. But in this case the macro lens is third-party and the camera has no way to know about the added objective. So the only way to remove CA would be matching up the RGB bands within the images as shot. I've never seen that approach work with anything other than carefully constructed calibration targets, so I'd be dumbfounded if the camera is doing it as a matter of course.ChrisLilley wrote:Just checking, since I am more familiar with Nikon than Canon - the in-camera JPEG engine provides no suppression of lateral CA at all?
Imaging at 5X -- a tale of two objectives
Moderators: rjlittlefield, ChrisR, Chris S., Pau
-
- Posts: 674
- Joined: Sat May 01, 2010 6:12 am
- Location: Nice, France (I'm British)
- rjlittlefield
- Site Admin
- Posts: 23564
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
-
- Posts: 674
- Joined: Sat May 01, 2010 6:12 am
- Location: Nice, France (I'm British)
There is a brochureware-level explanation on the Nikon site.
As I understand it, its a simple linear scaling of the red and blue channels relative to the green. This is less precise than the quadratic or cubic scaling used by converters like DxO, but has the advantage that the required scaling can be automatically determined from the photo; it requires no prior characterisation of the lens or the lens/body combination.
Edited to add:
Nikon describe the benefits and drawbacks thus:
I will see if I have a lens with bad lateral CA and develop a photo from raw with and without the correction.
CaptureNX2 also claims to reduce axial CA, in the latest versions. In my experience this is far less successful, and seems to be just a detection and desaturation of purple edges. Thankfully this is off by default in the raw converter and in the JPEG engine
As I understand it, its a simple linear scaling of the red and blue channels relative to the green. This is less precise than the quadratic or cubic scaling used by converters like DxO, but has the advantage that the required scaling can be automatically determined from the photo; it requires no prior characterisation of the lens or the lens/body combination.
Edited to add:
Nikon describe the benefits and drawbacks thus:
The mention of non-CPU lenses (those without an electronic connection between camera and lens, so the camera does not know which lens is connected) indicates that the correction does not depend on knowing lens characteristics. The mention of PC lenses (tilt/shift) shows that the only assumption is that the aberration is radially symmetrical about the image centre, which is clearly not the case for a tilt/shift lens.Moreover, because aberration is corrected regardless of lens type — including telephoto, wide-angle, fisheye, non-CPU and other types of NIKKOR lenses except special lenses such as PC lenses — this feature contributes to the sharper images professional photographers demand, from both Nikon FX and Nikon DX formats.
I will see if I have a lens with bad lateral CA and develop a photo from raw with and without the correction.
CaptureNX2 also claims to reduce axial CA, in the latest versions. In my experience this is far less successful, and seems to be just a detection and desaturation of purple edges. Thankfully this is off by default in the raw converter and in the JPEG engine
Last edited by ChrisLilley on Sat Jan 08, 2011 1:03 pm, edited 2 times in total.
I'd guess because a PC lens deliberately uses lens distortion to "correct" perspective, so any CA will not be constant / symmetrical across the field ?ChrisR wrote:This phrase struck me as odd, if it's a lens-independent function:
"other types of NIKKOR lenses except special lenses such as PC lenses ".
rgds, Andrew
"Is that an accurate dictionary ? Charlie Eppes
"Is that an accurate dictionary ? Charlie Eppes
- rjlittlefield
- Site Admin
- Posts: 23564
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
I've not run across anything like this from Canon. The Nikon brochure mentions specifically the D3 and D300 lines, but no others. Can I presume that it's not provided in the D5000 that I recently got?ChrisLilley wrote:There is a brochureware-level explanation on the Nikon site.
As I understand it, its a simple linear scaling of the red and blue channels relative to the green. This is less precise than the quadratic or cubic scaling used by converters like DxO, but has the advantage that the required scaling can be automatically determined from the photo; it requires no prior characterisation of the lens or the lens/body combination.
--Rik
- rjlittlefield
- Site Admin
- Posts: 23564
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
As I understand the PC lenses, they actually do correct perspective using tilt/shift and a distortion-free lens just like a view camera. But yes, the tilt/shift part certainly does mean that CA will not be symmetrical around the center of the frame.AndrewC wrote:I'd guess because a PC lens deliberately uses lens distortion to "correct" perspective, so any CA will not be constant / symmetrical across the field ?
--Rik
I hope that I am not adding to the uncertainty about what can be done through general purpose image processing, but my memory tells me that Photoshop Bridge (CS5, and perhaps some previous versions) also perform some CA removal. I can't verify this or test it further as my primary Photoshop PC is down for repairs. From memory I think I had a fair amount of luck improving some zoom telephoto photos (not macro photos) using this Photoshop component. One has a slider to adjust until a subjectively preferred image is obtained. I hope I'm not confusing this with some other tool. I've also successfully used the DxO image processing package to remove moderate amounts of this image defect, again based on my aging and definitely fallible memory.rjlittlefield wrote:Cool! I like getting dumbfounded occasionally, especially when it's for technological improvements of that caliber.
Do you have some references to articles describing how and how well it works?
--Rik
Is it possible that reasonably simple CA effects are similar enough that a generic CA image enhancement algorithm could make nice improvements in the sense of removing most of this type of degradation? One would not usually need to have mathematically correct CA correction, but just an improvement that is visible but without many introduced artifacts.
Yet I am also suspicious of such things as a matter of general experience. Usually one quickly runs into problems as one tries to make aggressive use of such corrections...
-Phil
"Diffraction never sleeps"
"Diffraction never sleeps"
-
- Posts: 674
- Joined: Sat May 01, 2010 6:12 am
- Location: Nice, France (I'm British)
Not sure. That brochure is a little old; lateral CA suppression is also in the D90 that I have, for example, and the D5000 is more recent.rjlittlefield wrote: I've not run across anything like this from Canon. The Nikon brochure mentions specifically the D3 and D300 lines, but no others. Can I presume that it's not provided in the D5000 that I recently got?
To check - take a photo in NEF+Fine JPEG. Then develop the NEF in some other software than NX2 (or, if you have NX2, switch off the axial CA supression). Compare the JPEG with the developed RAW.
Edit: or check a review. dpreview confirms that the D5000 has this feature:
Nikon D5000 vs D60: Key Differences:
Automatic correction of lateral chromatic aberration
-
- Posts: 674
- Joined: Sat May 01, 2010 6:12 am
- Location: Nice, France (I'm British)
Yes, as I mentioned above its simply a linear radial scaling of the R and B channels relative to G, centered on the middle of the image.DQE wrote: Is it possible that reasonably simple CA effects are similar enough that a generic CA image enhancement algorithm could make nice improvements in the sense of removing most of this type of degradation? One would not usually need to have mathematically correct CA correction, but just an improvement that is visible but without many introduced artifacts.
Obviously, this does not change the effects of lateral CA on a single channel. Nor does it help with lenses where lateral CA increases, then decreases, along the image diagonal.
- rjlittlefield
- Site Admin
- Posts: 23564
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
Great! I'll put that on the list of things to check out, time permitting. That'll be a while, though.ChrisLilley wrote:dpreview confirms that the D5000 has this feature:Nikon D5000 vs D60: Key Differences:
Automatic correction of lateral chromatic aberration
--Rik
- Joaquim F.
- Posts: 192
- Joined: Wed Apr 28, 2010 3:06 pm
- Location: Tarragona, Spain
- Contact: