What constitutes a 'correct' histogram?

[Manfred M]

Terminology is something that we can always split hairs over. I have the same issue as Urban when it comes to what is happening at the lower end, where the situation is effectively a lack of signal (due to a lack of photons), not an overflow situation. On the other hand, common use tends morph terminology, even though the original meaning of the definition gets garbled. If someone writes "clipping shadow values", most people will understand what is meant. In some ways this is a "feature" of the English language and allows it to evolve - there is no English "language police" that keeps the language "pure". I'm sure the Acadamie Francaise would certainly frown on this type of use in the French language and do its best to stamp it out.

With respect to under exposure, one of my professors, who was not sympathetic to students showing up with heavily underexposed work, would always say "just add more light". Just let me add, it wasn't a landscape photography course...

[Inkanyezi]

Maybe the thread can be wrought back to subject?

I think the question is a bit odd, as any histogram is a correct histogram. But maybe it was meant to ask how to get the "best" histogram, which is rather easy to answer.

There simply isn't a "best" histogram. When talking RAW histograms, ideally the whole histogram should fall within limits; i.e. if the histogram ends just at the baseline at both ends, all tones within the scene can be rendered in the resulting image.

The shape of the histogram can also tell us something. A histogram that consists of a hump, one and only one hump or maybe even a huge tower, then the scene is a low contrast scene. It is low key when the hump is to the left and high key if the hump is to the right. Nevertheless, a hump is typical of a low contrast scene.

A high contrast scene can have the entire histogram flat or rather flat, or it can have some smaller humps. If there are two humps to either side of the histogram, with a mostly flat histogram inbetween them, it is indeed a sign of a high contrast scene. But whichever way you put it, the resulting histogram depends on the scene and the exposure. If the entire histogram is within bounds, not piling up to any side, but sloping toward the little tips where it vanishes at each end, the dynamic range of the scene is contained within the exposure.

As a general rule, mostly we don't want the histogram to pile up at the very right edge, the brightest part of the image. Some clipping pixels can be allowed, but not wholesale clipping with an upward slope toward the right border.

The left end is different. sometimes it is simply inevitable that dark pixels, pitch black, pile up toward the left end of the histogram. This is OK, as long as important image information is not lost in total blackness. There are solutions if you need those dark areas to show more detail. Either you can just expose more, if there is space on the right side, with a lot of baseline to the right, which indicates that the image was perhaps underexposed. But, you can also do a HDR image by combining two or more exposures, merging them and tone mapping, to get an LDR image of the HDR scene. It is also possible, if the bright pixels are not important to image content, to just blow the highlights and expose for the shadows.

So the histogram may carry important information, and every histogram is 'correct', but if you have a lot of baseline without image content to the right side of it, you are free to use it by exposing more, thus getting less noise in the shadows and more latitude to pp manipulation.