p.2 #1 · 5d iii help: horrible over saturated reds.
Here's an example, I was in a rush to get these done so didn't have time to really look into fixing it. I will do though so I can sort it out easier in the future, so any tips gratefully received
p.2 #2 · 5d iii help: horrible over saturated reds.
The RAW file doesn't have a gamut assigned to it. It is just three maps of relative scene brightness as seen through red, green and blue filters. What the three maps are relative to is what you decide to expose at or above clipping in any of the three channels.
In ACR or LR when the RAW is assigned a "working" gamut for editing the 255 values of red, green and blue the camera recorded (if any) are assigned to the outer boundary the editing gamut you select. You'll have the same number of possible discrete colors in the file (a function of bit level) but they just get distributed differently. By way of analogy the colors are cows and the gamuts are different size pastures for them to graze in. The grass is always greener just over the fence and in PhoPhoto the fence is further from the barn than in the sRGB pasture.
If you want optimal highlight detail and subtle gradients in the highlights you don't want to blow any of the channels. For example skin, which reflects more red than green or blue will clip first in red, but you optimally want the red in the brightest highlight about 2/3 stops below clipping. For a solid white object exposure should be pegged 1/3 stop below clipping so the file value winds up around 245-250 at capture.
How to judge that when shooting? Use the clipping warning like you'd sight in a new gun. Take a shot, see where it lands, adjust your aim point. When shooting a portrait I increase exposure until I start to see clipping in any white clothing near the face then back down 1/3 stop. If they are not wearing any clothing that is white I raise exposure until the skin highlights are starting to clip then back off 2/3 stop. That works for ambient, flash and ambient and flash together.
In the shot above I exposed the sunny back of the jacket 1/3 stop below the clipping warning in Av with - EC and the front 2/3 stop below clipping warning with flash FEC.
and on white backgrounds... http://super.nova.org/MP/WhiteBGTowelCard2.jpg
With studio lighting I raise fill until I see the camera capture shadow detail above the noise threshold first. Next I set the brightest light hitting the subject (rim light above) 1/3 stop below clipping warning. To be perceptually correct as in the outdoor shot the frontal "key" must be slightly darker. I raise it last adjusting it 2/3 stop below clipping. The result? A full range of "seen by eye detail" everywhere, which is both the technical and perceptual definition of optimal exposure.
I rely on the camera metering to get the exposure "in the ball park" but having used TTL metering for years I know it doesn't and will never "guess" what I want correctly exposed more than half the time so I engage my eyes and brain to evaluate and adjust from the camera metering's "I think this is right" baseline.
Primary color RGB content in scenes in bright lighting will cause that respective channel to clip. Whether you, who ultimately control exposure, decides to let it clip or not needs to be determined by understanding the scene contrast, the DR limit of your sensor, how scene vs sensor DR is affecting the exposure of midtones and shadows and what needs to retain detail in the shot.
If for example you are shooting a person in backlight standing next to a stop sign in sunlight the stop sign will clip in the red channel long before red channel in the shaded face is normally exposed 2/3 stop below clipping. What is the "correct" exposure in that situation? Depends on what is most important, the subtle texture in the stop sign or a normal view of the face? You can't have both, and the camera can't decide for you. The camera metering in a situation will usually hedge it's bets and expose somewhere in the middle, still blowing the detail in the red sign but also underexposing the face by a stop or more at EC=0. You, looking at that EC=0 shot from the camera's "I think this is right" baseline must engage your eyeballs, brain and creative mojo and decide to shift it - 2 to preserve the detail on the sign or +2 to make the face normal. Or you can put the camera on a tripod, shoot it both ways and blend the two in Photoshop.
At the point where photgraphers accept electonic finders and there's no flapping mirror slowing down frame capture rate it will be possible to do in-camera HDR like that in many situations to solve the dilemma of scene exceeding recording range that has existed since the 40's when color film came into wide use. Been waiting patiently for that, but at this point just hope I live long enough to actually see it. Until then I never venture outside without a flash on my camera so when the scene > sensor I can change the scene range, at least in the foreground, to fit the sensor perfectly as is trival in a studio setting with flash.
p.2 #3 · 5d iii help: horrible over saturated reds.
The RGB capture process has inherent limitations. For example yellow, cyan and magenta objects are very difficult to capture as seen by eye because in all of them two of the three channels will be maxed out and carrying very little detail. The eye senses differently, more like an Lab file with two types of cone cells sensitive to blue/yellow and green/magenta in the center 2° of the FOV and green sensitive rods covering the rest of the retina. The rods are 3000x more sensitive than the cones which is why the green region is larger in color spaces and why image with a lot of green in them are tiring to look at and play tricks with color perception.
Much of the technical side of photography and reproduction deals with working around the limitation of the RGB color separation process which has been around for over 100 years and is well understood. There are Photoshop tricks that can restore detail the eye can see in yellow flowers but the RGB camera can't record I illustrated recently here: https://www.fredmiranda.com/forum/topic/1116199/0#10657624
It fakes the detail in the missing channel by copying detail from the channel in the RGB file which has detail into the L channel of any Lab copy of the file, increasing detail without altering color. The Lab copy is then coverted back to RGB and blended into the original with masking in the flowers.
p.2 #4 · 5d iii help: horrible over saturated reds.
I encounter this problem sometimes when dealing with red in general, but a certain shade of red (that matches the bayer pattern red in the camera) is the worst as capturing that is the worst case for the camera. If there is no green or blue channel component at that area on the bayer grid, it means the camera cannot do the interpolation it normally does, meaning a third the resolution and more noise.
Even with normal reds that are a slight off shade from the 'true red' I still get this general problem, so when shooting something very colorful I typically use -1/-2/0 bracketing or something like that, and then sometimes end up picking the -1 or -2 EV shot and balancing the colors from there. This ends up looking better than the 0EV shot even though the 0EV shot may not be overexposed on the red channel according to the histogram (camera or raw development tool). It seems like when it is nearing overexposure, the steps between one level of red and the next start to get noticeable so the key is to somewhat underexpose the reds a bit and then map them up to make it look good, different from the ordinary practice of 'expose to the right'.
I can get a similar phenomenon when dealing with blues, but not quite as bad or pronounced as the reds.
And this seems inherent with just about every digital camera out there, not just Canon. People I know who use Nikon, Sony, or whatever else get the same thing.
p.2 #5 · 5d iii help: horrible over saturated reds.
Excellent.
Yes, this is a "problem" - though I'd prefer to think of it as a reality - of digital sensors and exposure. It is all too easy to blow the red channel, even when shooting raw, if you don't look at the multi-channel histogram. It does, indeed, mean "underexposing" by a stop or two when one channel is very hot in your subject. This channel is typically the red channel. We've all seen countless photographs of subjects like sunset illuminated clouds, mountain peaks, flowers, and more that are clearly blown out. I see it sometimes even in the work of some well-known photographers who might know better.
Exposure has always been tricker than people want it to be. While your meter could have given you a simple "correct" exposure with film, sophisticated film photographers understood (and still understand) that the actual content of the scene will often call for a very different exposure and very different methods of chemical post-processing, otherwise known as "developing."
The same is true with digital. While we might like to think that if we just expose the way the camera "says" to, that the exposure should be "right," subjects are still just as variable and complex today as they were when we shot film.
The issue of color spaces is a different thing that is related to what you do with the raw file, and not to what is in it. If you blow out a channel in the camera, you aren't going to get it back in post. (Sometimes you can sort of fake it, but you are still faking it.)
Dan
ausemmao wrote:
As the others have said, I don't think it's a problem unique to the 5D3 - My D7000 and D3100 do it and the D700 and D800s I've tried do as well - it's due to the amplification applied to the signal. If the red channel is properly clipping, the only thing I know of that works is to go into 3 channel histogram at exposure time and make sure red doesn't clip, which usually involves a stop or more of "underexposure".
With that said, what skibum is referring to is that the profiles that LR applies can provoke clipping where there is none in the raw file. Setting the LR colourspace to prophoto RGB and then manipulating it in there could mitigate the problem.
Hahahahahahhaaha. I wish that were true. When people talk about not liking Nikon skin tones, it's usually because they've unknowingly run into this. The orange/yellow shift it creates is not pleasant. ...Show more →
p.2 #6 · 5d iii help: horrible over saturated reds.
Your train photo is probably more or less a perfect storm of challenges in regards to blowing the red channel. First, the problem areas is perhaps the very brightest portion of the scene. Second, the subject in the most shaded area is painted black!
I'm having a hard time even figuring out what I'm seeing on the yellow/red area at the front of the engine. Was subject originally all yellow? And did the shadows turn red? That is bizarre. I can't say that seen anything quite like that before.
This might be essentially an impossible scene to shoot in natural light and a single exposure.
p.2 #8 · 5d iii help: horrible over saturated reds.
morganb4 wrote:
No it was red, thats what Im saying, it was as red as the connecting rod, the horn and the buffer plate of the adjacent engine in the bottom left.
I metered the scene and knocked it back a bit ti keep the shadows.
There are a few other images like this too.
If so, that looks to me like some kind of weird bug somewhere in the capture or processing chain. That's not just an example of an oversaturated red; it's blown beyond all recovery.
Does the image look this way processed in different RAW converters? If using Adobe, have you tried different processes to see if they render the same (2003, 2010, 2012)?
p.2 #9 · 5d iii help: horrible over saturated reds.
Just looking, not measuring colors in PS, I'd guess that the red of the horn and connecting rod has a slight green component (i.e. it's a slightly orangey red). So, I'd speculate that the very bright front surface has clipped the red significantly, while allowing the green to show greater effect, resulting in yellow. The shadows are enough dimmer that the red hasn't clipped badly, so they are close in hue to the connecting rod, etc.
gdanmitchell wrote:
Your train photo is probably more or less a perfect storm of challenges in regards to blowing the red channel. First, the problem areas is perhaps the very brightest portion of the scene. Second, the subject in the most shaded area is painted black!
I'm having a hard time even figuring out what I'm seeing on the yellow/red area at the front of the engine. Was subject originally all yellow? And did the shadows turn red? That is bizarre. I can't say that seen anything quite like that before.
This might be essentially an impossible scene to shoot in natural light and a single exposure.
p.2 #10 · 5d iii help: horrible over saturated reds.
I encountered something similar with UV with my first digital camera in 2001 (Kodak DC290) when used with Vivitar 285HV flash. Blue objects like jeans and my Mac G4 case would be rendered cyan in the photo. Turned out that the camera didn't have a very good UV cut filter over the sensor and UV brighteners in the jeans and plastic registered like additional visible light on the sensor. I concluded that after profilng the camera (I had the tools and know-how because I ran a printing center) and confirmed it with an RIT professor who was using the same camera to teach students how to profile one (I found him with a net search). Curiously the problem didn't occur in daylight or with the built-in flash, which it turns out had a UV filter over it. The Vivitar 285HV must produce a lot of UV,
In this case since it is reds in sunlight that are going wonky there may be a problem with the IR cut filter on the camera. Maybe the IR filter inserter robot was on a coffee break when your camera was built.
p.2 #12 · 5d iii help: horrible over saturated reds.
morganb4 wrote:
No it was red, thats what Im saying, it was as red as the connecting rod, the horn and the buffer plate of the adjacent engine in the bottom left.
I metered the scene and knocked it back a bit ti keep the shadows.
There are a few other images like this too.
I've seen this before, once you blow out the reds, then you start to get those "wierd" color shifts due to the other channels blowing out later than the red channel does. Typically in my case it's red shifting to magenta-ish, in your case yellow. Which is far more obvious, yellow is pretty unusual. But like one of the other posters said, if there is some green component there like a brownish-red that's typically used in railroads, that's probably what you are seeing.
Do you by any chance have different exposures of one of these scenes like the one in your photo?
Notice how when the front of the train is in the shadow of those protrusions, it is closer to it's actual color (red) rather than yellow.
It's just the nature of the beast, I could post plenty of examples but they are much more subtle, that is much more obvious one.
Other than "don't overexpose", another solution here is to use a color filter on the camera to re-balance the colors (ie. removing half the red amplitude while leaving the other colors mostly untouched) and then removing the color shift that the filter creates in post.
p.2 #14 · 5d iii help: horrible over saturated reds.
hard to say, can you post up the RAW?
maybe you exposed it too much and simply blew out the red there so badly it turned to yellow (a little green becomes a lotta green since red is already maxed = yellow), if you hadn't your other detail would be lost, anyway that is why you need 3 more stops DR in this case you could, if you had it with you, tripod and take two exposures for HDR, since it looks like nothing is changing in the scene at all
looks like the raw itself is blown (it is for sure unless there is still room for some extreme highlight rescue, although if it has gone that far yellow in a basic develop it's doubtful it can be retrieved) and that it might be beyond sRGB gamut anyway, maybe
although the gamut thing appears to be the least of the issues and minor compared to the over-exposure, most likely, bit hard to say without seeing the RAW
p.2 #15 · 5d iii help: horrible over saturated reds.
It looks like the front of the train is illuminated by a strong light (based on the shadows), and the black train is in the shadows. The effect is definitely quite strange, but as was already pointed out, it looks like this was the perfect storm of issues.
p.2 #16 · 5d iii help: horrible over saturated reds.
morganb4 wrote:
No it was red, thats what Im saying, it was as red as the connecting rod, the horn and the buffer plate of the adjacent engine in the bottom left.
I metered the scene and knocked it back a bit ti keep the shadows.
There are a few other images like this too.
You may have metered "the scene," but that doesn't mean that all elements in the scene will fall with in the averaged exposure that the camera gives you. That is what I was writing about in my earlier post.
Do you have your camera set up to show the RGB histogram and to show the "blinkies" that indicate blown out areas? If not, turn those on.
When your camera is probably doing something like center weighted metering for auto exposure. This gives preference to the center of the frame but "acknowledges" areas outside of that. The grossly bright area of the front of the train - both a very "hot" color and the brightest area in the frame - is likely outside of the area that was metered. So your camera was trying to give you and exposure for some of the much darker areas of the scene and, not unexpectedly, the colorful areas got completely blown out.
Again, some scenes simply cannot be captured in single unadjusted shot. I would probably have kept knocking down the exposure to avoid blowing the brightest area, knowing full well that I would have to do my best to bring back some shadow detail in post. Or, if using a tripod, I would have made more than one exposure (exposure bracketing) at different settings and combined them in post.
In any case, it is virtually always a bad idea to let a large area like this one blow out, even if that means that you need to use a setting that might seem like underexposure.
p.2 #17 · 5d iii help: horrible over saturated reds.
Photon wrote:
Just looking, not measuring colors in PS, I'd guess that the red of the horn and connecting rod has a slight green component (i.e. it's a slightly orangey red). So, I'd speculate that the very bright front surface has clipped the red significantly, while allowing the green to show greater effect, resulting in yellow. The shadows are enough dimmer that the red hasn't clipped badly, so they are close in hue to the connecting rod, etc.
That is a pretty good description of why colors shift when you blow one color channel. The image color is a balance of the three color channels. When you push one of them past it upper boundary, it cannot record any more luminosity for that color. If the color is red and the you go to where you max out the red channel, the green and blue channels will have much less luminosity. Let's say that they have, for example "20%" luminosity at the point when the red channel is at what I'll call 100% luminosity. If you push beyond that, the red channel will still be at 100%, but the luminosity of the other channels will continue to increase, let's say to 30%. Obviously, if the "real" color was 100, 20, 20, then 100, 30, 30 is going to look like a different color.
p.2 #19 · 5d iii help: horrible over saturated reds.
I think in the train shot the camera has done it's job by exposing to record detail in the very dark side of the train but as a consequence the front is wildly overexposed which has led to the bleeding of the reds, it's just more than the sensor can handle. Really it needs a reduction in contrast by a flash fill or whatever.
If the shot is important I'd develop the RAW file twice at different exposures and combine them.
p.2 #20 · 5d iii help: horrible over saturated reds.
morganb4 wrote:
Thanks all but I would like to reiterate that in 7 years of shooting digital (RAW) from a 1D2N a 1Ds2 a 1D3 a 5D a 1D4 and now the 5D3 plus a rash of compacts, this is the FIRST time I have seen this.
You can see on other aspects of the train and the train just off to camera left, how the red should look.
I'm curious about the lighting, I see the big halogens on the ceiling. What's behind your back? Open air?
