What is the reason for the color fringing at high contrast edges on the CP950 and will it be fixed on the CP990?

From the sample pictures I've seen, the CP950 has purple fringing and the CP990 has red/brown fringing.

-- Drury Woodson (drury@woodson.net), March 19, 2000

Answers

Mr. Boznian, (who responded earlier) I take issue with the response that only Nikon suffers-I have seen this in most 2.11 megapixel cameras and I believe it is a combination of lens and CCD technology in high contrast areas. Outside of using a different aperture (not wide open) or using the lower contrast feature (a Nikon exclusive) it is a problem to deal with in after capture manipulation.

Ignore what you see on the 990 samples-that is a CCD noise-not fringing.

Michael

-- Mike Rubin, Product Marketing Manager Nikon (web@imaging-resource.com), March 23, 2000.

I've seen this on my Oly 2000C. I sent an example photo to Olympus. I enclose extracts from the two replies that I got:

-------------------------------------------------------

After speaking with our technical department in Germany, it would seem that the effect you experienced with the bright, almost ethereal line, running down the left-hand side of your picture, is something that they had seen before. It apparently is not a fault of the camera, more a limitaton of current CCD technology in digital cameras (other manufacturers included.) It transpires that when a digital camera is faced with exteme contrast, ie a shaded wall against a very bright sky, then because the CCD cannot record the full range of tones (much like a conventional film would struggle), it tries to interpolate, ie guess, what the dividing line between the two areas of extreme contrast should be. The only way around this with a digital camera, would be to try to avoid taking photos where there is massive contrast. The other, more convoluted method, would be to purchase a CLA-1 lens adaptor for your camera and a 43-55mm step-up ring (both Olympus products available from PHOTOGRAPHIC stockists) and then to employ some sort of filter system and to use a grey graduated or split neutral density filter, (which is common practice when photographers using conventional cameras, are faced with scenes with extreme contrast.) This effectively reduces contrast in the scene and would help the camera enourmously. Your photographic dealer would be able to offer further advice.

-------------------------------------------------------

The line is caused by a to high contrast in the picture were the algorithms of the camera in connection with a over controled CCD shows an artefact. To avoid this problem please do not take pictures in such a strong back light situation.

-------------------------------------------------------

They didn't say that it would be *fixed* in future cameras, I notice...

-- Mark Steven Davis (glod@hotmail.com), March 19, 2000.

That business about CCD artifacts that you should try to fix with a filter of some kind is ridiculous. "Lateral color", a particular kind of chromatic aberration, is an aspect of lens quality. It takes a lot of careful design and manufacture to make a lens that does not produce color fringing at the edges. Canon and Sony and almost everyone but Nikon/Coolpix lately has been able to make lenses without noticeable color fringing. The CCD artifact color someone else described above would be EVERYWHERE in the image where there was harsh contrast, not just at the edges of the picture frame, since CCD's work the same at their edges as they do at their centers. Color fringing at picture edges is a LENS problem, and we are all wondering when Nikon, which "clearly" has the ability to do so, is going to fix it.

-- Russell Bozian (finaldesign@hotmail.com), March 19, 2000.

Well, I'll come down on the side of this one that the "color fringing" problem Drury is referring to affected a number of the 2 megapixel cameras out there. (It was also called the "purple fringe" or "purple haze" (in a nod to Jimi Hendrix) problem.) There's been a lot of discussion about it on the 'net, and it does seem to be more of a CCD effect than a lens one. The reason I say this is that a number of people pointed out that the orientation of this particular defect wasn't radially symmetric, which you'd expect from a lens problem. It also doesn't appear to be "blooming" on the CCD, as some have suggested. One theory I heard was that it has to do with the three- dimensional structure of the surface of the CCD, and some sort of shadowing of the active elements by other structures on the CCD surface. Another suggested it was a refraction problem in the microlenses on the surface of the CCD.

Whatever the case, it was very different from chromatic aberration (if I'm thinking about the same thing Drury is referring to), and definitely seemed to be associated with cameras using the 2MP Sony CCD chip.

For an example of what it looks like (an extreme case), visit the following album on ClubPhoto, which was shot with an Olympus C2000.

http://ir.clubphoto.com/david72827/Blowing_Rock_NC_Blowing_Rock_NC/

Look at the top left corner of the "Cone Path" image, where the leaves/ branches are silhouetted against the sky, and you'll see the bright purple haze that I believe plagued all the 2MP cameras using the Sony chip. (In this picture, it looks more or less radially symmetric, but I've seen other shots where it doesn't appear so.)

Don't know that this will clarify anything, but at least it demonstrates that this isn't exclusively a Nikon problem. (FWIW, I believe the Canon S10 also shows the same behavior, although I don't have an example picture at hand to demonstrate.)

-- Dave Etchells (detchells@imaging-resource.com), March 23, 2000.

Dave,

It *IS* radially symmetric. There's literally thousands of examples of the effect to prove it too. The effect is never visible in the exact center of the image. As you move away from the center in any direction, the fringes grow in size. If you image a black object against a white background (black ink on white paper is enough) you'll find a purple (a bit redder than purple on the cp990) fringe on the sides of the object toward the center of the image, and a greenish fringing on the sides of the object away from the center of the image.

The images at this link:

http://photo.askey.net/forums/read.asp?forum=1007&message=126034

show it quite clearly. They're from the upper left of the Imaging Resource resolution target images of both the cp950 and cp990. If you crop from the lower right corner the fringing would be reversed. From the right, it would be purple on the left side, and green on the right.

It is totally consistent with color chromatic aberration, where the lens simply does not focus all the wavelengths of light onto the same spot. The problem with declaring it to be chromatic aberation and going home, is that it generally can only been seen with high contrast edges, and the lens shouldn't care if the light is bright or not. And again, the pixels don't have to be blown for it to happen. They weren't in either of the crops shown above. Just black ink on white paper. Maybe it's happening throughout the image but you just can't see it as easily when the luminance values are much lower. But there's also no reason to believe that an engineer would design the camera to handle CCD to RGB image interpolation in a radial pattern.

Some wild hypothesises exist which could help explain. Perhaps CCD sensors don't handle light coming in from an angle as efficiently. Maybe there are reflections off the edges of one sensor onto its neighbor. This stuff is a bit out there.

Peter iNova has done some interesting experiments with decreasing the size of the red channel centered in the middle of the image (as close to radial as you can get with a resample), which can have a dramatic effect on the appearance, generally for the better, but really isn't a cure for all photos.

What I can't understand is how people are still debating whether the effect EVEN EXISTS.. Sheesh. You'd think a bunch of photographers could believe photographic evidence when presented to them.

Oh, and Mike, this we've seen the low light noise and stuff in the early CP990 samples. This isn't it. The fringing shown above is 100% consistent with the fringing seen in the CP950 and with that in other digicams. Denying the existence of the problem, or suggesting that a shot of black ink on white paper, when the paper isn't even overexposed, is too extreme for the camera, isn't going to make us go away.

Don't get me wrong. I fully intend to buy a CP990, but I'd like to be able to put a technical name behind what anyone can see and reproduce. I'm an engineer. I just want to know what it is. I already know how to fix it for my own photos in post processing. Giving this a real name will at least stop the speculation.

ian

-- Ian Frechette (frechett@houseofthud.com), March 24, 2000.

I have noticed this even on middle quality flatbeds like the Agfa DuoScans but the new Agfa Fotolook 3.5 software upgrade seems to have diminished the CYANish/Green fringing. Anyone out there with a Duoscan have the same experience? Anyway I hope this doesn't happen on a Nikon D1. I would think that a $5000 camera (plus $360 for the lens) would have a much more superior image quality than a $999 coolpix. It would be a major disappointment - I guess I'll find out soon, I ordered one w/ a 60mm micro. I'm also curious about the Fuji S1 which also uses nikkor lenses is coming out this summer. The Super CCD sample image looks interpollated and doesn't look impressive at all.

-- Bert C. (bert@longlivethemac.com), March 28, 2000.