Expose to the right written 2 years ago

I happened upon an article today about maximising the S/N ratio in digital photography — entitled Expose Right. The basis of this article was that one should endevour to overexpose an image as much as possible — but without causing any clipping, as this would give an image with the best tonal range once normalised.

1 2

1. An underexposed histogram; 2. An overexposed histogram

I advise reading the article if you want the following to make a whiff of sense.

Now to me, those histograms look to contain the same amount of data, but Mr Michael Reichmann suggests that the over-exposed image contains 16 times the amount of information. His reason being that a camera’s sensor is linear, and light is exponential. Reichmann’s article makes much more sense on this, so I won’t try to explain here.

Reading this I thought it had to be made up, and so embarked on a Sunday evening of exploring this wild claim to see if it was true.

The experiment consisted of taking two photos of the same subject (some delightful woodchip wallpaper), one overexposed by 2.5 stops, and one underexposed by 2.5 stops. Each time attempting to put the histogram in the upper/lower fifth, without causing clipping. The images would then be “normalized” in photoshop — to give us a properly exposed image.

3 4

3. The underexposed image, once normalised; 4. The overexposed image, once normalised, both shown with before and after histograms. (Click for big)

From the images it’s easy to see that the underexposed image has lots more noise, and lacks the smooth transitions that the over-exposed image has. I was really surprised to see how pronounced the effect was, but no matter how I fiddled, I couldn’t get the under-exposed image to behave anything like the over-exposed image.

So it seems it’s correct wot the man said. Woo.

← previous entry | next entry → Sun 27th Jan 2008 - 22:25 | no comments | tagged with Photography