Why the problems with red?

I've been noticing that red, especially red cars, are extremely hard to photograph well. Does it have to do with the Bayer pattern only having 25% red? If so, I would think the problem would appear with blue as well, though. I've been noticing this on most photos of red vehicles I see, not just mine.

Here's a (not so good) picture I took of my truck a while back, and all the shots have this spotted red look, like amplified noise. I don't remember what ISO I shot this at, but I think it was low:

Byte size: ZERO | Content warning: NOT AN IMAGE

It might be because the red is blown out. Since you're using a 20D - with a luminance histogram, not an RGB histogram like the Mark II - you wouldn't see it when chimping the histogram. It wouldn't be the Bayer pattern because then blue would do the same since blue is also 25%. Try shooting something very red and then move your EC down 1/3 and see if it looks the same in both shots. Or expand PS's histogram to RGB and see if the Red is blown out in the original.

I don't think it's a color space issue - I might if it were yellow blowing out, but the reds are pretty well covered in both Adobe RGB and sRGB. But I'm not positive. I did review a couple of UncleDoug's profile posts and the reds look full on all profiles he showed.

Phil,

A couple of quick observations.

Use your color picker and set the info palette to HSB (hue, saturation, brightness). In this scheme, pure red=0 (on a scale of 0 to 360 degrees). Your trucks values are around 354. Brightness and saturation are greater than 80% - thus a very vivid red! There is a "correct" hue value for this truck - but we don't really know what it is. The hue should stay nearly constant as you sample different areas of the truck under the same light, even as saturation and brightness vary.

Because the red channel is clipped, you have lost red tonal data - thus the "spots" you are seeing represent posterization - lack of smooth gradations within the red tones of the image.

Varying the white balance of the image doesn't much influence the red color of the truck. Color balance is not main problem.

The problem is intrinsic to the profile used to convert from the camera color space to the final color space - sRGB in this case. This is the type of color (saturated reds) that color gurus argue on and on about which RAW converter does a better job (ACR vs Capture One, etc). Capture One has the flexibility to use alternative profiles. In PS ACR, you would have to work with the ACR calibration tab settings to modify the default profile - very difficult to do well.

A easier fix is to use a "Selective Color" adjustment layer. Work on the "reds" and adjust the magenta slider. You probably can get very close to what you think is right.

Bottom line - to get this type of image "right" straight out of the camera, you would have to create a custom profile for your camera under the lighting condition used. This not usually realistic. Thus you would use profiles made by others and/or make post-processing color corrections with Hue/Saturation or Selective Color. Capturing in Adobe RGB will give a little more room to avoid clipping of the red channel.\

Good lookin' truck

I think it's been mentioned here that since Red is the R of RGB and with a very red object you end up with single channel blowout. Sky that is very blue can end up being very, very solid blue with no color variation.

I've been b1tching about this for quite a while and nobody cares.

It's all the fault of luminous landscape "Expose to the right" article. 99% of the cameras have only the luminosity histogram, so pushing luminosity to the right GUARANTEES blowout of separate color channels. Too many people take this rule too literally, which results in shots that look flat and unappealing.

JAF Doorhof never has this problem and his shots are not only saturated, but have detail in them... unless he overexposes for effect. The man uses a lightmeter.

What is the solution? Underexpose by 1.5 stops. Your luminosity histogram is going to move a bit to the right, but the information in the individual channels will be captured.

Hope this helps somewhat

I have run into this problem with both reds and blues, which causes me to think that it is a bayer sourced artifact. I notice it most when I convert to mono and push the red channel.

Regardless of the source, I am able to work around it by selcting the area (The truck, in this case) and then doing a combination of Gaussian and Smart blurs only in the red channel only in the selected area. It does not make it perfect, but improves it quite a bit.

Also, I would not be suprised if this comes from the blue channel. Since there is so little blue in the truck, that that is there might take on a blotchy appearance.

Rad

It has nothing to do with bayer. If there's enough green to blow out the green channel, then it's going to be blown out even if there's 2x photosites on the sensor.

But you do get less green noise in your camera at the higher ISO because there's 2x the amount of green cells.

Underexpose, and your histogram will move to the left.

---Bob Gross---

I have some success (depending on how badly the channel(s) is clipped) using the Linear response curve in Capture One, then applying my own curve.

Yes, that's what I meant. Being ambidextrous I always had problems with left and right.

You should see me giving directions while somebody's driving - turn right (2 lane changes)... I mean left... left...

Thanks everyone. These are all great suggestions. I do see some nearly perfect examples of red cars, so I know it's mostly me, but I do see a large percentage that look like what I posted above.

I should have mentioned my example was with the Zenitar 16mm, so I don't even remember what aperture I had used, etc. I'll give it a go with the 10-22 soon and see if I can do any better using some of the techniques mentioned here.

Or, maybe I should just go with the 1 series so I will have the multi-channel histogram.

And Doc, I did read that Luminous Landscape article a while back and I usually try to push the histogram to the right. I'll bet that's a big part of the problem with just the luminance histrogram on the 20D in this case.