Histograms :-s

[Katy Noelle]

How did we manage to get from post #1 with Morse's side-kick in Oxford asking a simple question about histograms, to this...?

(Note: Morse is a famous detective character from a TV series of that name in the 1986-2000 period. And I should know as I have just watched all 33 episodes on Love films.)

It's been brewing!

[_GUI_]

After seeing so many photographers taking the histogram as the Truth to validate their images, I wrote this article trying to explain some myths about the histogram.

Unfortunately it's in Spanish (online translation available), but will try to summarize:

THE HISTOGRAM. THAT UNEXPECTED GUEST

The histogram is a summarization of the information contained in our image. So it must always be considered together with the image, never alone. For instance in the histogram all spatial information is lost. The following two images have exactly the same histogram (you can check):



The histogram is very useful, and can be used for many things, but not for everything.


CORRECT USES OF THE HISTOGRAM

1. CHECK EXPOSURE

Over exposure:


Under exposure:


Black point clipping:



2. COLOUR MANAGEMENT

After a colour profile conversion, colours falling out of the destination gamut get clipped producing peaks in both ends of the histogram. Peaks on the right end could be avoided through some underexposure. But peaks in the left end mean the output colour profile cannot support some colours and there is not a solution for this without altering the colours (e.g. apply desaturation):




3. UNSHARP MASK

Strong sharpening typically leads some pixels to get saturated or clipped to 0. This is seen in the histogram, but it's an expected effect and we don't need to worry about it in the histogram, but looking for oversharpening in the image:




4. COLOUR CASTS

A local histogram (PS allows for this) can provide more accurate information about colour casts and how much we are cancelling them than looking at the image itself:

[_GUI_]

WRONG OR MISLEADING USES OF THE HISTOGRAM

1. THE BETTER THE HISTOGRAM'S SHAPE, THE BETTER THE IMAGE

There is not a standard perfect histogram. The goal of image edition is to get the best image, not a perfect histogram.

Some claim the histogram should have the shape of a symmetric bell curve and reach both ends. Wrong, it will depend on the image. Look at your image at the same time you look at its histogram.








2. THE MIDDLE GRAY IS 128 IN THE HISTOGRAM

Wrong, middle gray (let's assume Lab Luminance 50%) will have a different value according to the gamma of the colour profile used, so don't expect to get 128 printed as a middle gray on a correctly calibrated system. E.g. in ProPhoto RGB middle gray is value 100:




3. THE HISTOGRAM AND COLOUR PROFILES

The histogram strongly depends on the colour profile used, so any tool working on levels of the histogram (Levels, Curves,... in RGB mode) will have a very different effect if applied in a different colour profile.

Wider profiles are much more sensitive to tonal arrangements since we need to apply softer changes to get a visible effect:




4. THE HISTOGRAM AND IQ

Some claim a full histogram means high IQ while a combed histogram means low IQ. It is usually like that, but this can also be a totally wrong assumption. To find out about IQ look at your image!.

These two images have the same IQ:


These two images have lower IQ than the former with the posterized histogram:

[jiro]

Thank you very much for your very detailed examples and explanation, Guillermo. I really appreciate your hard work preparing this material and sharing it to us. Thanks!

[rob marshall]

I thought this was posted here, but it's in another thread. So here it is (if you haven't already seen it)...