Color: is Nikon superior to Canon ?

[luong]

I saw on the blog of LLoyd Chambers (http://diglloyd.com/diglloyd/2008-05-blog.html#_20080505ColorRendition) the following entry:

"Having been shooting the 21MP Canon EOS 1Ds Mark III for the past 3 weeks while the D3 was away, I had gotten used to the dullness of the 1DsM3 images (well, not really). Examining the D3 images this evening, I was once again struck by the dramatic difference in the vibrancy in both color and tone. Viewing D3 images is like viewing the original, looking at the 1DsM3 images is like looking at a copy."

My own experience with the Canon  1DsII and III is that in natural landscape photography the colors need a considerable amount of adjustment before they are as pleasing as those from slide scans. In other words, I'd agree that the color of Canon is not very good out of the box. However, I have no experience with current Nikon cameras.

If you use both current Canon and Nikon cameras, has your experience has been similar ?

Tuan.

[BernardLanguillier]

If you use both current Canon and Nikon cameras, has your experience has been similar ?
[a href=\"index.php?act=findpost&pid=212708\"][{POST_SNAPBACK}][/a]

The question is probably whether this quote refers to jpg straight out of the camera, or whether it is about a conversion from raw.

If it is the latter, then the main thing is probably that the default rendition of the converter used by LC is set to produce more vivid colors for the D3.

There is a real possibility that one sensor is able to better capture colors than the other, but as far as I know nobody has taken the time to compare this for colors/levels of illumination other than those related to standard color charts under normal illuminations.

It is generally said that Nikon colors tend to be a bit more yellow, while Canon tends to be a bit more red.

I am personnally happy with the colors I get from my D3, but there are plenty of Canon users out there that are happy about the colors they are getting from their DSLRs.

Cheers,
Bernard

[madmanchan]

Lloyd's comments are essentially about what Nikon/Canon do to colors when interpreting raw files. Or put another way, what do Nikon/Canon do when making JPEGs? (As I recall, Lloyd is using NX to develop the D3 files.)

There's nothing inherent to any of the modern Canon or Nikon DSLRs regarding capturable colors that makes one superior to another. The only differences are in the default raw conversion settings (or "looks")  being presented by the manufacturers.

[michael]

I'd add to what Eric says, that using DNH Profile Editor along with Lightroom 2 or ACR 4.5, there is now essentially no need for there to be differences between the way two cameras interpret colour.

In fact, if you want a Nikon to look like a Canon, or visa versa, you can easily do this. It's a new ball game.

Michael

[Slough]

Whilst having no experience of modern Canon cameras, I do wonder whether the lenses played a role in what was seen. Nikon lenses are known for giving contrasty results, which I saw recently when using a Tamron 90mm macro. The lower contrast of the Tamron lens gave the images a flat appearance, making it less natural and less real, which I did not like. I've also seen this with spotting scopes, where Nikon and Leica have a contrasty natural look, whereas the Swarovski and Zeiss image are faltter and IMO less 'real'.

[bjanes]

The question is probably whether this quote refers to jpg straight out of the camera, or whether it is about a conversion from raw.

If it is the latter, then the main thing is probably that the default rendition of the converter used by LC is set to produce more vivid colors for the D3.

There is a real possibility that one sensor is able to better capture colors than the other, but as far as I know nobody has taken the time to compare this for colors/levels of illumination other than those related to standard color charts under normal illuminations.

It is generally said that Nikon colors tend to be a bit more yellow, while Canon tends to be a bit more red.

I am personnally happy with the colors I get from my D3, but there are plenty of Canon users out there that are happy about the colors they are getting from their DSLRs.

Cheers,
Bernard
[{POST_SNAPBACK}][/a]

Bernard,

I have no idea of the relative accuracy of colors between Nikon and Canon, but there are theoretical reasons why there could be differences. As mentioned previously, some of these differences lie in the differences in processing of the raw data, either in camera or with a raw converter.

This [a href=\"http://billjanes1.home.comcast.net/~billjanes1/binary/KopieSensorChar.pdf]paper[/url] by Kopie concerns Foveon sensors, but the same principles apply to cameras with Bayer arrays. A matrix conversion is usually done to transform the camera color space to a working space such as XYZ. The matrix coefficients depend on the spectral qualities of the RGB filters of the Bayer array. The filters may not have a spectral response similar to the response of the RGB sensors in the eye and this can lead to problems. However, if the filters have a linear response, it is possible to produce an exact metameric match with the matrix conversion (stated by Thomas Knoll on the Adobe ACR forum).

In the real world no such linear filters exist and the coefficients are usually chosen to produce the best overall results. Some colors may match and others will be off, and it is possible to calculate a metamerism index as shown in the paper. It would be easy to derive such an index from results with a Macbeth ColorChecker. If you used a ColorChecker DC, which has more colors, the process would be more involved. As you mention, results with real world images in the field are not readily available.

As I understand the calibration process using the Fors script and similar methods, one is merely tweaking the coefficients used for the matrix conversion. Still some colors will be off, and one may gain additional accuracy by using a lookup table and this process is used by the Adobe ProfileEditor.

Regards,

Bill

[Graeme Nattress]

Yes, you calibrate by finding a matrix that converts the raw sensor data to XYZ. However, there is never one "perfect" matrix to do so. You start by going for a matrix that produces least errors delta E ab in Lab space, and tweak from there for it to look good.

Graeme

[bjanes]

Whilst having no experience of modern Canon cameras, I do wonder whether the lenses played a role in what was seen. Nikon lenses are known for giving contrasty results, which I saw recently when using a Tamron 90mm macro. The lower contrast of the Tamron lens gave the images a flat appearance, making it less natural and less real, which I did not like. I've also seen this with spotting scopes, where Nikon and Leica have a contrasty natural look, whereas the Swarovski and Leica image are faltter and IMO less 'real'.
[a href=\"index.php?act=findpost&pid=212744\"][{POST_SNAPBACK}][/a]

Yes, there is a tradeoff between contrast and resolution in lens design. It has been said that Zeiss lenses are designed for resolution and Leitz lenses for contrast, but I don't know if this is true or how Nikon and Canon design their lenses.

Resolution above Nyquist is undesirable in digital cameras is undesirable since it produces aliasing artifacts. Our Nikons and Canons have blur filters to lessen the artifacts. It is better to have good contrast below Nyquist.

Bill

[bjanes]

Yes, you calibrate by finding a matrix that converts the raw sensor data to XYZ. However, there is never one "perfect" matrix to do so. You start by going for a matrix that produces least errors delta E ab in Lab space, and tweak from there for it to look good.

Graeme
[{POST_SNAPBACK}][/a]

The Delta E ab is widely used but isn't that accurate for comparing visual differences, as [a href=\"http://www.imatest.com/docs/colorcheck_ref.html#colorerr]Norman Koren[/url] explains. Differences in luminance and saturation contribute to Delta E ab, but most viewers perceive these differences as less significant than than changes in hue. In fact, a small Delta E ab is not even desired by many photographers who prefer more saturation.

Memory colors such as skin tones, blue sky, and green foliage are particularly important. Back in film days many photographers used Velvia for landscapes, but Caucasian skin tones can be terrible with this film. However, I have read that National Geographic photographers often found African skin tones pleasing with this film. The new camera raw has default profiles for landscape and portraits, which is a major advance IMHO.

Bill

[digitaldog]

Bernard,
I have no idea of the relative accuracy of colors between Nikon and Canon

Neither are accurate in the context of this discussion. Both are attempting to produce pleasing JEPGs which is totally subjective. What's more accurate, Velvia or Ekatacrhome? Neither really.

The use of the term accurate color in digital photography is mostly marketing hype. There IS accurate color one can achieve, its usually not at all pleasing to view:


http://www.color.org/ICC_white_paper_20_Di...ment_basics.pdf

[bjanes]

Neither are accurate in the context of this discussion. Both are attempting to produce pleasing JEPGs which is totally subjective. What's more accurate, Velvia or Ekatacrhome? Neither really.
[{POST_SNAPBACK}][/a]

That is correct, but with the new PE and ACR one can tweak the colors to individual preference, depending on the subject material. Landscape and portrait photography require different parameters.  It is possible that one camera would yield better results for the intended purpose than another and this could be qualified by a metamerism index.

With regard to color accuracy of films, it is often said that Kodak strove more for an accurate neutral appearance while Fuji favored more saturation. As you point out, accurate is not necessarily most pleasing and Kodak had to bring out its own color enhanced films.

The use of the term accurate color in digital photography is mostly marketing hype. There IS accurate color one can achieve, its usually not at all pleasing to view:
[a href=\"http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf]http://www.color.org/ICC_white_paper_20_Di...ment_basics.pdf[/url]
[{POST_SNAPBACK}][/a]

For some purposes such as reproducing fine art paintings of such masters as Rubens and Monet, accuracy is desirable. For some scientific and documentary work, accuracy is also desirable but one might want to increase color saturation or alter the tone curve to bring out small differences.

That link is a good one. [a href=\"http://wwwimages.adobe.com/www.adobe.com/products/photoshop/family/prophotographer/pdfs/pscs3_renderprint.pdf]Karl Lang[/url] discusses this topic in more detail.

[Chris_Brown]

Neither are accurate in the context of this discussion. Both are attempting to produce pleasing JEPGs which is totally subjective. What's more accurate, Velvia or Ekatacrhome? Neither really.
[a href=\"index.php?act=findpost&pid=212774\"][{POST_SNAPBACK}][/a]

In other words, capture the data and process it to your own liking. If perfect colorimetric accuracy is required then it's only derived from one's own, custom, condition-centric, camera profile. Everyone else's camera profiles will render different results. Which result will be perfectly, colorimetrically accurate? Will it look good? Will the client buy it?

[digitaldog]

In other words, capture the data and process it to your own liking. If perfect colorimetric accuracy is required then it's only derived from one's own, custom, condition-centric, camera profile. Everyone else's camera profiles will render different results. Which result will be perfectly, colorimetrically accurate? Will it look good? Will the client buy it?
[a href=\"index.php?act=findpost&pid=212789\"][{POST_SNAPBACK}][/a]

Perfect colorimetric accuracy (EVEN with copy work) would mean the measured colors of the scene are the result and that's almost always NOT going to make a desired visual match let alone a pleasing one.

The only way we can really define an accurate colorimetric value is to measure it at the scene. If you want to produce that out the back end (to display or print), fine. Its accurate. Does it appear as you saw it? Is it pleasing? Probably not in either case.

The ICC white paper describes this well if I do say so myself. Accurate color is scene referred. But scene referred imager needs to be output referred so there goes the statement with respect to accuracy.

[digitaldog]

For some purposes such as reproducing fine art paintings of such masters as Rubens and Monet, accuracy is desirable.

No, color matching is desired. That doesn't mean, and in probably all cases doesn't mean colorimetric values in equal colorimetric values out.

Considering the enormous dynamic range and gamut disconnect between the input and the output, the values pretty much have to change and thus become output referred which while producing what everyone could say is a color match, isn't accurate in terms of absolute colorimetry. And if that's not the case, how can we gauge accuracy?

[bjanes]

No, color matching is desired. That doesn't mean, and in probably all cases doesn't mean colorimetric values in equal colorimetric values out.

Considering the enormous dynamic range and gamut disconnect between the input and the output, the values pretty much have to change and thus become output referred which while producing what everyone could say is a color match, isn't accurate in terms of absolute colorimetry. And if that's not the case, how can we gauge accuracy?
[a href=\"index.php?act=findpost&pid=212805\"][{POST_SNAPBACK}][/a]

Is not color matching one measure of accuracy? In the case of reproducing a painting, there is not much difference in dynamic range--both are reflective and have a dynamic range of 100 or 200 to 1. If the dynamic range of the input and output are dramatically different, then luminoisty would have to be compressed with a tone curve, but one would still want the hues to be similar. Of course, luminosity does affect perceived hue: As Dan Margulis points out yellow at high luminosity is perceived as such, but is perceived as brown at low luminosity.

[digitaldog]

Are not accuracy and color matching the same thing?

Not in my book. First off, two different people could disagree upon color matching. So how do you decide? You measure. Accurate color is measured color IMHO. That's the idea behind colorimetry. If the numbers have to differ to produce a match, it becomes difficult to define accuracy.

That isn't to say one could not use math and measurements through a CMS to say that the resulting values accuracy transform from original measured data to new colorimetric values. But that works with solid colors, not at all as well or at all, with images which are composed of thousands or millions of such solid colors.

Colorimetry with solid colors is relatively easy to handle. Saying two solid color samples match, either visually or colorimetrically  (which is what today's CMS's are based upon) isn't difficult. Saying two images or a scene and an image match using the same process doesn't work well. Maybe when we have better color appearance model based CMS's that can be accomplished. But its important to remember that the CMS we use today is based on the notion of two solid colors matching and then extrapolating that to a heck of a lot more such colors we call an image.

This is just one reason why the current systems fail in a number of areas (like the fact that calibrating a display to D50 and viewing images on such a device don't match a print viewed under D50).

[digitaldog]

In the case of reproducing a painting, there is not much difference in dynamic range--both are reflective and have a dynamic range of 100 or 200 to 1.

Do you think there's a difference printing the same image to an Epson using Archival Matt and Premium glossy? Same inks and device, only difference is paper. It can be a significant difference in just this one example yes?

Of course, luminosity does affect perceived hue: As Dan Margulis points out yellow at high luminosity is perceived as such, but is perceived as brown at low luminosity.

So where does that put Dan and this apparent issue with appearance in terms of accuracy and colorimetry (something I suspect he's not tuned to in either case)?

[bjanes]

Do you think there's a difference printing the same image to an Epson using Archival Matt and Premium glossy? Same inks and device, only difference is paper. It can be a significant difference in just this one example yes?
[a href=\"index.php?act=findpost&pid=212817\"][{POST_SNAPBACK}][/a]

One of the main differences between the papers is that the glossy has a higher DMax. If you wanted to reproduce a Monet as well as possible, you would probably choose a paper that best matched the properties of the original canvas. My point is that perfect color matching is not currently possible, that there may be variations among cameras, and there are objective methods by with color matching can be measured. Ultimately, color is a perception and not something that exists in the scene. The measurement method that best correlates with the perception of multiple subjects would be the preferred method.

So where does that put Dan and this apparent issue with appearance in terms of accuracy and colorimetry (something I suspect he's not tuned to in either case)?
[a href=\"index.php?act=findpost&pid=212817\"][{POST_SNAPBACK}][/a]

Not invented here complex?

[digitaldog]

One of the main differences between the papers is that the glossy has a higher DMax. If you wanted to reproduce a Monet as well as possible, you would probably choose a paper that best matched the properties of the original canvas.

Agreed.

[Graeme Nattress]

The Delta E ab is widely used but isn't that accurate for comparing visual differences, as Norman Koren explains. Differences in luminance and saturation contribute to Delta E ab, but most viewers perceive these differences as less significant than than changes in hue. In fact, a small Delta E ab is not even desired by many photographers who prefer more saturation.

Memory colors such as skin tones, blue sky, and green foliage are particularly important. Back in film days many photographers used Velvia for landscapes, but Caucasian skin tones can be terrible with this film. However, I have read that National Geographic photographers often found African skin tones pleasing with this film. The new camera raw has default profiles for landscape and portraits, which is a major advance IMHO.

Bill
[a href=\"index.php?act=findpost&pid=212765\"][{POST_SNAPBACK}][/a]

Yes, that's just the start of the problems.... Hue errors are perceived as much more an issue than saturation errors, and luma errors are very hard to spot. However, Delta E ab does serve as a very good starting point to begin tweaking from. It is indeed common practice it seems to alter the saturation to pump up the colours, and make the image pop. And yes, when calculating matrices, often some colours accuracy is weighted more than others so as to get "important" colours more accurate at the expense of others. It's a fun game to play, and quite fraught with subtle details.....

Graeme