Little White Lies

Cambridge In Colour has a couple of truly excellent articles on histograms (here and here).  I'm not going to go into detail on histograms other than to say that a) they're tools for understanding the distribution of brightness/darkness in an image, and b) understanding histograms is absolutely required for anyone interested in advancing beyond "point and shoot" in digital photography.

Anyway, the CIC articles treat the topic in quite a bit more detail than you generally find on the 'Net or in consumer-oriented books.  What I found so insightful is how they demonstrate the limitations of histograms, and the common RGB histogram in particular.  To make a long story short (for the long story, read the articles), RGB histograms can give you "false positives".  In certain, not uncommon, situations they can indicate that you're blowing out or severely underexposing photos, when you're actually not.  It has to do with the fact that the "all channel" histogram depicts the area in which the individual color histograms would overlap if they were all drawn onto the same graph - without regard for whether or not the overlapped values come from the same pixels (sounds confusing, but read the articles and the examples will make it really clear).  The bottom-line is:
  • When you have an image with areas of multiple primary colors (red, green, and blue) at near-maximum intensity, an RGB histogram is likely to show blowouts even though there may be no white anywhere in the image.  Imagine taking a photo of a field of red tulips in spring time under a deep blue sky...
  • When you have an image with areas of color that have no primary color contribution, an RGB histogram is likely to show severe underexposure (a mountain falling off the left side of the histogram) even if there's no black in the photo.
Now, this doesn't mean that one shouldn't use RGB histograms!  As I said, they're absolutely required.  But knowing their limitations is extremely helpful if you want to avoid pulling your hair out because your histogram is telling you you've got a big problem and you simply don't see it.