Home · Register · Search · View Winners · Software · Hosting · Software · Join Upload & Sell

Moderated by: Fred Miranda
Username   Password

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Canon Forum | Join Upload & Sell

1      
2
       3       end
  

Archive 2012 · 5dmk2, problems with pure reds (and sometimes blues)
  
 
gdanmitchell
Offline
• • • • •
Upload & Sell: On
p.2 #1 · p.2 #1 · 5dmk2, problems with pure reds (and sometimes blues)


First, I have seen it with green - though, it is rare. The problem most often manifests in the red channel. It is not limited to the 5D2, so anyone thinking that this is an issue related to a specific camera is barking up the wrong tree.

You are on the right track when you note that some images or image elements might provide luminosity almost entirely to a single color channel. The issue is simply that this channel gets "swamped" (i.e. - over exposed) even though the overall image luminosity seems relatively low. Although this isn't strictly an accurate way to describe it, it is essentially that you overexpose the red channel and underexpose the others. In the same way that highlights (say clouds in the sky) blow out when you over expose an image, here the reds blow out - and in the same way that cloud highlight detail is lost and you get pure white, red highlight detail is lost and you get essentially pure red.

The notion that the camera should always be able to correctly interpret and expose for any complex situation that is thrown at it, without intervention or judgment by the photographer is unfortunate. Cameras can do this in many, many cases, but sometimes we still need to make some decisions ourselves when confronted by challenging lighting and other situations. Here, I have learned that when the scene is very "hot" in one color channel that a) there is a risk of blowing out that channel, b) the averaged luminosity histogram display will not tell you this since it averages the three color channels, c) it is a great idea to use the RGB histogram display the separately shows the three color channels, and d) you should, in most cases, err on the side of underexposure rather than overexposure.

I believe that current sensors still have twice as many green sensing photo sites. However, I think your explanation, while it has some interesting merit, is not getting at the essential problem or the relatively simple solution to it.

Dan

Access wrote:
Actually I was thinking about this late last night and thought of something, is this what is going on?

When the part of the photo is pure red, that means it has no green or blue component. Meaning it will only show up on the red part of the bayer array, and not the green and blue part which will be zero or close to zero. Meaning, one third (or less if the bayer doesn't favor red) the advertised resolution, and less luminence data to average over or interpolate from. The raw conversion software does the best job it can, but
...Show more



Feb 03, 2012 at 04:34 PM
alundeb
Offline
• • • •
Upload & Sell: Off
p.2 #2 · p.2 #2 · 5dmk2, problems with pure reds (and sometimes blues)


Chrominance noise reduction can go wild with the reds. Make sure it is set to 0 at low ISO.



Feb 03, 2012 at 05:42 PM
Access
Offline
• • •
Upload & Sell: Off
p.2 #3 · p.2 #3 · 5dmk2, problems with pure reds (and sometimes blues)


gdanmitchell wrote:
You are on the right track when you note that some images or image elements might provide luminosity almost entirely to a single color channel. The issue is simply that this channel gets "swamped" (i.e. - over exposed) even though the overall image luminosity seems relatively low. Although this isn't strictly an accurate way to describe it, it is essentially that you overexpose the red channel and underexpose the others. In the same way that highlights (say clouds in the sky) blow out when you over expose an image, here the reds blow out - and in the same way
...Show more
No I've definately seen this too. For instance when taking photos under 2700K or 2500K lighting while relying on the camera's automatic metering. But it is slightly diffferent issue? It can happen with any shade of red, not just a pure one. You get an obviously blown out red channel, and often a wierd color shift for part(s) of the photo where the red was overblown -- when you try to apply the white balance correction. As it's trying to bring the clipped red channel back in. At this point, I typically just switch to manual mode to get a proper exposure on the red channel.

It seems to me like the worst case for a bayer array sensor is a pure color that matches a color of the bayer array, since that will not appear (or appear as black) on the other pixels. I've seen this same problem with my panasonic camera too, only it's much more pronounced (less bits coming out of that sensor?). I've tried bracketing like in the screenshots I posted and the brighter exposures will look a little better, but the problem is still there.



Feb 03, 2012 at 06:04 PM
Photon
Online
• • • • • •
Upload & Sell: On
p.2 #4 · p.2 #4 · 5dmk2, problems with pure reds (and sometimes blues)


Pixel Perfect wrote:
Blues have been a problem since film days. Certain deep blue/violet/indigo colours are very hard to capture accurately.

I would think this is always a problem for digital RGB because indigo and violet are shorter wavelength than primary blue. Our eyes (i.e., our brains) interpret the lack of response from green receptors (which do respond weakly to pure blue light) as indicating hues beyond blue. Not easy to implement that with monitors, as they only have RGB pixels.

As for the OP's red issues, I'm convinced that it's a combination of the factors already mentioned: misleading histograms because of WB, only 1/4 of the sensels dedicated to red, and the lack of "assistance" from color contrast when attempting to record tonal gradations in an area of pure hue. The practical solution is also clearly what has been recommended - weaker exposures and/or bracketing, and use of post processing when the subject matter calls for it in order to render the look that was perceived by the photographer at the time of exposure.



Feb 03, 2012 at 07:04 PM
snapsy
Online
• • • • •
Upload & Sell: On
p.2 #5 · p.2 #5 · 5dmk2, problems with pure reds (and sometimes blues)


Access wrote:
No I've definately seen this too. For instance when taking photos under 2700K or 2500K lighting while relying on the camera's automatic metering. But it is slightly diffferent issue? It can happen with any shade of red, not just a pure one. You get an obviously blown out red channel, and often a wierd color shift for part(s) of the photo where the red was overblown -- when you try to apply the white balance correction. As it's trying to bring the clipped red channel back in. At this point, I typically just switch to manual mode to get a
...Show more

There are many possible causes to color issues but I'm not exactly clear what aspect of the IQ you're specifically having issues with. It is true that resolution for the red/blue channels is materially below green, both because it's sampled less but also because the exposure/sensitivity for these channels lag the green channel.

Regarding the 5DM2 in particular, its color selectivity of the red channel is far below it's peers, including the 5D. This means that to produce red hues for 5DM2 images, the raw converter must add extra gain to the red channel while reducing gain from the green channel. This causes several side effects, the most prominent being extra noise and thus less detail when you do color post-processing. You can use DxoMark to see the differences in color selectivity between sensors. Here are the pages for the 5D and 5DM2 for instance:

http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Canon/EOS-5D
http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Canon/EOS-5D-Mark-II

Click on the 'measurements' tab, then click on 'color response'. The white balance matrices show what the raw processor has to do to produce the given primary (go from a 'raw' color channel to a post-gamma RGB representation). Pick the primary on the left hand side of the matrix, then look across to the RGB columns to its right to see what positive/negative boosting needs to be done to produce the color.

For the 5D, red is produced via the following: R+1.96, G-1.09, Blue+0.13
For the 5DM2, red is produced via the following: R+2.25, G-1.5, Blue+0.25

See how much more red must be boosted and green reduced to produce red on the 5DM2 vs the 5D? This relates to how selective the sensor is to color, meaning how well it can distinguish both different colors and shades of a given color. A great resource for understanding this is DxoMark's case study of the D5000 vs T2i, on the "color blindness and sensor quality page". Here's the link:

http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Canon-500D-T1i-vs.-Nikon-D5000/Color-blindness-sensor-quality



Feb 03, 2012 at 07:06 PM
gdanmitchell
Offline
• • • • •
Upload & Sell: On
p.2 #6 · p.2 #6 · 5dmk2, problems with pure reds (and sometimes blues)


Yes, it does not need to be pure red - just a color that is hot in the red channel. It happens with orange California poppies. It happens with yellow and other colors of fall foliage. I've seen it happen with green in images that consisted almost entirely of green foliage.

If you think about it, the color shifts are the logical result of the blown out channel. The overall color is defined by the balance among the three color channels in the sensor. But if one maxes out and cannot record any stronger levels but the others do not max out, you essentially create a balance that is, overall, less balanced toward the blown color. Let's say that you have some color that is red:256, green:100, and blue:100. Now over expose it a bit and you get blue:150, green:150, but still only red:256 - and the balance among the three colors is distorted.

Dan


Access wrote:
No I've definately seen this too. For instance when taking photos under 2700K or 2500K lighting while relying on the camera's automatic metering. But it is slightly diffferent issue? It can happen with any shade of red, not just a pure one. You get an obviously blown out red channel, and often a wierd color shift for part(s) of the photo where the red was overblown -- when you try to apply the white balance correction. As it's trying to bring the clipped red channel back in. At this point, I typically just switch to manual mode to get a
...Show more



Feb 04, 2012 at 03:33 AM
gdanmitchell
Offline
• • • • •
Upload & Sell: On
p.2 #7 · p.2 #7 · 5dmk2, problems with pure reds (and sometimes blues)


The problem here is that the theoretical musings ("being extra noise and thus less detail when you do color post-processing") are not born out in actual practice. I have shot tens of thousands of frames on a 5D and also on a 5D2, and post processed raw images from both in just about every way possible... and what you predict simply doesn't turn out to be true.

We are trying to make this way too complicated. If the OP simply reduces exposure a bit to avoid blowing out the red channel the problem will almost certainly be resolved. This may require some post-processing adjustments to bring up the darker area of the image, or some alternative like exposure blending. And this will be the case with a 5D or a 5D2 or just about any other DSLR making such a photograph.

Dan

snapsy wrote:
There are many possible causes to color issues but I'm not exactly clear what aspect of the IQ you're specifically having issues with. It is true that resolution for the red/blue channels is materially below green, both because it's sampled less but also because the exposure/sensitivity for these channels lag the green channel.

Regarding the 5DM2 in particular, its color selectivity of the red channel is far below it's peers, including the 5D. This means that to produce red hues for 5DM2 images, the raw converter must add extra gain to the red channel while reducing gain from the green channel.
...Show more



Feb 04, 2012 at 03:37 AM
snapsy
Online
• • • • •
Upload & Sell: On
p.2 #8 · p.2 #8 · 5dmk2, problems with pure reds (and sometimes blues)


gdanmitchell wrote:
The problem here is that the theoretical musings ("being extra noise and thus less detail when you do color post-processing") are not born out in actual practice. I have shot tens of thousands of frames on a 5D and also on a 5D2, and post processed raw images from both in just about every way possible... and what you predict simply doesn't turn out to be true.

We are trying to make this way too complicated. If the OP simply reduces exposure a bit to avoid blowing out the red channel the problem will almost certainly be resolved. This may require
...Show more

I agree, his issue may simply be hue shifts from a clipped channel. Regarding the theoretical musings and whether color selectivity has actual observable (and detrimental) effect on IQ for PP, my experience is quite different than yours but I'll table that discussion since in may not be pertinent to the OP's issue.



Feb 04, 2012 at 05:27 AM
trumpet_guy
Offline
• • • •
Upload & Sell: Off
p.2 #9 · p.2 #9 · 5dmk2, problems with pure reds (and sometimes blues)


Just a couple of shots for comparison from a 30D (Canon glass, then Zeiss glass):

http://tswen.smugmug.com/Nature/Flora-and-Fauna/IMG2230CLONED/755872345_Ssg4u-X2.jpg


http://tswen.smugmug.com/Nature/Flora-and-Fauna/IMG6550/755873557_cpJs4-XL.jpg

I find the reds very easy to blow out on Canon DSLRs. I don't have experience with
Nikon to compare.

Try shooting well illuminated red objects with varying exposure to see where red clipping
occurs.

I would recommend using sRGB color space unless you are experience with other
color spaces. sRGB has a narrower gamut than aRGB, which should make tonal
gradations smoother, but will limit how deeply saturated red you can display. I
think for your experiment, sRGB is preferred.

Tim



Feb 04, 2012 at 05:37 AM
skibum5
Online
• • • • • •
Upload & Sell: Off
p.2 #10 · p.2 #10 · 5dmk2, problems with pure reds (and sometimes blues)


trumpet_guy wrote:
I would recommend using sRGB color space unless you are experience with other
color spaces. sRGB has a narrower gamut than aRGB, which should make tonal
gradations smoother, but will limit how deeply saturated red you can display. I
think for your experiment, sRGB is preferred.

Tim



no way, look at those roses developed in prophotorgb 16bit on a wide gamut monitor and then look at them in sRGB 16bit, sRGB kills deep reds like in roses



Feb 04, 2012 at 06:19 AM
 

Search in Used Dept. 



trumpet_guy
Offline
• • • •
Upload & Sell: Off
p.2 #11 · p.2 #11 · 5dmk2, problems with pure reds (and sometimes blues)


If he has a wide gamut monitor, I expect you are right.



Feb 04, 2012 at 06:22 AM
skibum5
Online
• • • • • •
Upload & Sell: Off
p.2 #12 · p.2 #12 · 5dmk2, problems with pure reds (and sometimes blues)


I'm suspicious that they can be born out in practice, although I haven't carefully tested so I'm not sure.

And they just might be relevant to some of his issues his since he mentions that sometimes turning HTP OFF makes it a bit better, while HTP should make things better if we were just talking pure red channel clipping, so it sounds like something he may have issues in deeply shaded reds way down in the shadows perhaps? If so, then snappy's stuff might be a bit relevant. Again, why would simply turning HTP off ever help him if all of his red issues are just from bright clipping? I'm sure many are but it sounds like often it's more than just that.


gdanmitchell wrote:
The problem here is that the theoretical musings ("being extra noise and thus less detail when you do color post-processing") are not born out in actual practice. I have shot tens of thousands of frames on a 5D and also on a 5D2, and post processed raw images from both in just about every way possible... and what you predict simply doesn't turn out to be true.

We are trying to make this way too complicated. If the OP simply reduces exposure a bit to avoid blowing out the red channel the problem will almost certainly be resolved. This may require
...Show more



Feb 04, 2012 at 06:23 AM
crazeazn
Offline
• • •
Upload & Sell: Off
p.2 #13 · p.2 #13 · 5dmk2, problems with pure reds (and sometimes blues)


its just a canon + red thing.


Feb 04, 2012 at 07:12 AM
DownEaster
Offline

Upload & Sell: Off
p.2 #14 · p.2 #14 · 5dmk2, problems with pure reds (and sometimes blues)


Coming from Olympus to Canon this is one of the first things I noticed. Actually I noticed even before I made the switch. I never noticed this issue with Olympus but immediately was confronted with it on my 7D.


Feb 04, 2012 at 07:38 AM
Access
Offline
• • •
Upload & Sell: Off
p.2 #15 · p.2 #15 · 5dmk2, problems with pure reds (and sometimes blues)


gdanmitchell wrote:
We are trying to make this way too complicated. If the OP simply reduces exposure a bit to avoid blowing out the red channel the problem will almost certainly be resolved. This may require some post-processing adjustments to bring up the darker area of the image, or some alternative like exposure blending. And this will be the case with a 5D or a 5D2 or just about any other DSLR making such a photograph.

Okay I spent some more time with this, playing with the raws and the bracketed shots from that day and one other test shot I did of a nearly pure red sweatshirt. And I think I see what you are saying. In some cases I have to underexpose by as much as 4 EV to get a proper exposure that does not blow out the reds.

But when I get that proper exposure, I can then apply the post I want and things look okay. Sure if I go down to the pixel level, I can still spot the artifact with the pure red parts of the photo, but that's not a big deal to me. It must be minimized enough to the point where it's not a problem when viewed at more normal resolutions.

Thanks all, I really do appreciate the help.



Feb 04, 2012 at 09:09 AM
snapsy
Online
• • • • •
Upload & Sell: On
p.2 #16 · p.2 #16 · 5dmk2, problems with pure reds (and sometimes blues)


Access wrote:
Okay I spent some more time with this, playing with the raws and the bracketed shots from that day and one other test shot I did of a nearly pure red sweatshirt. And I think I see what you are saying. In some cases I have to underexpose by as much as 4 EV to get a proper exposure that does not blow out the reds.

But when I get that proper exposure, I can then apply the post I want and things look okay. Sure if I go down to the pixel level, I can still spot the artifact with
...Show more

Another alternative is to use a color filter (cyan in this case), which change (balance) the temperature of light and let you keep a higher exposure without blowing a channel.



Feb 04, 2012 at 03:09 PM
gdanmitchell
Offline
• • • • •
Upload & Sell: On
p.2 #17 · p.2 #17 · 5dmk2, problems with pure reds (and sometimes blues)


skibum5 wrote:
I'm suspicious that they can be born out in practice, although I haven't carefully tested so I'm not sure.


I have a feeling that you might perhaps be saying something interesting, although I haven't read it carefully enough to understand, so I really cannot say... ;-)

crazeazn wrote:
its just a canon + red thing.


Nope. It is a DSLR plus red (or any hot color channel) thing. :-)

Access wrote:
Okay I spent some more time with this, playing with the raws and the bracketed shots from that day and one other test shot I did of a nearly pure red sweatshirt. And I think I see what you are saying. In some cases I have to underexpose by as much as 4 EV to get a proper exposure that does not blow out the reds.

But when I get that proper exposure, I can then apply the post I want and things look okay. Sure if I go down to the pixel level, I can still spot the artifact with
...Show more

You are welcome. I have had a lot of experience with this, and it doesn't create problems for me anymore when I consider the nature of the scene I'm photographing... which has always been necessary in photography.

In many cases, the camera (almost any modern camera) can do a creditable job of automatically coming up with an exposure that is "close enough for jazz," and sometimes it can do just as well as a smart, experienced photographer making manual adjustments. But when the scene presents complex exposure challenges (huge dynamic range, unbalanced overall color, multiple light sources of different temperatures, etc.) there is no AE system that can automatically make the "right" choice, much less the best choice. A smart and experienced photographer exercising experience-based good judgment can still outsmart the camera in such cases.

It is kind of pointless to blame the camera...

Another alternative is to use a color filter (cyan in this case), which change (balance) the temperature of light and let you keep a higher exposure without blowing a channel.

I've read that suggestion before, but I don't think it is going to work quite the way you imagine... unless you are going to be happy with the very "distorted" color balance that results. If you do want "natural" color - e.g. - very hot in one channel as in the original scene - once you compensate for the filter in post you'll be more or less back where you would have been had you simply avoided blowing out the hot channel at the time of exposure.

Take care,

Dan


Edited on Feb 04, 2012 at 03:36 PM · View previous versions



Feb 04, 2012 at 03:23 PM
crazeazn
Offline
• • •
Upload & Sell: Off
p.2 #18 · p.2 #18 · 5dmk2, problems with pure reds (and sometimes blues)


gdanmitchell wrote:
I have a feeling that you are saying something interesting, although I haven't read it carefully enough to understand, so I really cannot say... ;-)

Nope. It is a DSLR plus red (or any hot color channel) thing. :-)

You are welcome. I have had a lot of experience with this, and it doesn't create problems for me anymore when I consider the nature of the scene I'm photographing... which has always been necessary in photography.

In many cases, the camera (almost any modern camera) can do a creditable job of automatically coming up with an exposure that is "close enough for
...Show more

It def. is a hot color channel thing but the Nikon's seem to handle it better or perhaps its the glass.



Feb 04, 2012 at 03:35 PM
gdanmitchell
Offline
• • • • •
Upload & Sell: On
p.2 #19 · p.2 #19 · 5dmk2, problems with pure reds (and sometimes blues)


crazeazn wrote:
It def. is a hot color channel thing but the Nikon's seem to handle it better or perhaps its the glass.


Glass has nothing at all to do with this.

The "brand mythology" stuff is generally not too useful. You'll have this same issue with Nikon or Canon or Pentax or Sony or Leaf or Fuji or Olympus or you name it. It is simply the way sensors operate on digital cameras. (Actually, you can have the issue with film, though it will manifest in different ways.)

You may think you are seeing less or more of it on one camera than another because the camera software may be doing some of the very same kinds of compensation that I have described or, more likely, you have not done head-to-head comparisons but are instead relying on a sort of general feeling that the grass must be greener on the other side of the fence... or should I say, the color balance should be greener on the other brand? ;-)

Dan


Edited on Feb 04, 2012 at 03:43 PM · View previous versions



Feb 04, 2012 at 03:41 PM
CW100
Offline
• • • •
Upload & Sell: Off
p.2 #20 · p.2 #20 · 5dmk2, problems with pure reds (and sometimes blues)


crazeazn wrote:
It def. is a hot color channel thing but the Nikon's seem to handle it better or perhaps its the glass.


I think Nikon is actually worse with red





Feb 04, 2012 at 03:41 PM
1      
2
       3       end




FM Forums | Canon Forum | Join Upload & Sell

1      
2
       3       end
    
 

You are not logged in. Login or Register

Username   Password    Reset password