Histogram Colors doesn’t match point samples

[Wienke]

I’m looking at an RGB-mode image with a histogram showing nothing but red from brightness levels 220 to 255. That is, the green and blue channels show no info, zero output, from 220-255. Yet any point sample in those bright areas will show, in addition to the expected heavy red, also a moderate amount of green and a small amount of blue—that is, I get RGB values like 194-147-91. Understanding that, in RGB, red and green make yellow, and adding blue to them makes the yellow duller and brighter, I would expect that color to be a light tan. Indeed, that is what it is. So the histogram seems to be incorrect. Clearly I’m misreading the histogram, but how?

[D Fosse · Accepted solution]

Think about it. No G or B values above 220 - and that's exactly what you have (as per your example) - you have 147 and 91. Not above 220, right?

Nothing wrong here.

[Wienke]

Thanks for taking the time to look at this.

I am thinking of the 220-255, along the X axis of the histogram, as representing specifically High Brightness. Prompted by your reply, I do note that the RGB shape of the histogram is not the same as the Colors (all-colors superimposed) shape, so perhaps the individual color channel graphs use a different system. But if I use Curves to cut the Red channel off above 222, the bright areas do go greener, which seems to confirm my understanding that the X axis represents Brightness in all channels?

Put another way, I understand the values of 147 and 91 to represent the amount of green and blue, not brightness; they're the GB of RGB. Looking at the point sample expressed as HSB, its brightness is 91%. 91% of 255 is level 232. Which is what I would expect, because I sampled a very bright area of the image.

Put one more way, I don't understand why I'm not seeing ANY green or blue information along the brightest part of the X axis where I know those colors have to be present.

[D Fosse]

Forget about the Luminosity histogram. It just uses a formula to mix the three channels (roughly 25%-60%-15%).

The Colors histogram gives the real distribution of pixel values.

There is absolutely nothing unexpected in what you're describing. This probably becomes easier to wrap your head around if you go into the Channels panel and look at the channels individually. You'll see that the red channel has bright values all the way up, while the green and blue channels are drab and dark with no real bright values. As the histogram shows.

[Wienke]

I think I get it now. The X axis represents brightness ONLY for the RGB channel. For the individual color channels, X represents something else--which is where I've been confused, because the word "value" as an art term actually means brightness! But so, for the Red channel, the X axis represents a range from max cyan to max red. So that part of the All-Colors histogram where there is nothing but red is the part of the picture where there are no extreme blues, no extreme greens--or, better (yes, thanks, that's the clincher!), no pixels with G or B measures above 220!

(I think I'll leave the Luminosity histogram as terra incognita for the time being.)

So I should use the individual channels in Curves pretty much exclusively to control hue (and saturation); for brightness I should always be working with the RGB channel.

Thank you very much for your time. I really was stumped!