What's red and black and out of focus?

This entry is an expanded version of a response to a query from a photographer made in the discussion forum of the Ohio Valley Camera Club.  The shooter wanted to know why she couldn't get good results using autofocus shooting a red object on a black background.

Most of the answers had to do with distance from the object, but there's more to it than that.  Part of the problem is the way autofocus works (passive autofocus, at least, which is the kind you get on good digital cameras).  And part of it is the mysterious kinship between red and black.

Autofocus works by detecting an object in the frame and measuring the degree of contrast between it and something next to it.  It now makes a small change and sees if the contrast has become greater.  When it's at its greatest, the autofocus says "voila!" and stops adjusting.  If it can't find sufficient contrast, it keeps hunting or gives up in disgust.

The more contrast, the easier it is for the autofocus to do its job.  In scenes of low light, there's very little contrast.  In scenes with a monotone subject, such as snow on the ground or clear blue sky, it's also hard for autofocus to find enough contrast to grab a target. Same reason; not enough contrast for the autofocus sensor to gab hold of.

You can read a basic explanation of the autofocus process in How Stuff Works and a better one here (Cambridge Colour.com)

Now remember that we were talking about a red object on a black background, such as this one:

This image consists of pure red, exactly half way between full light and full dark (128 on a scale of 0-256). That's represented by the thin red line on the histogram show upper right in this screen shot. No problem for autofocus, right?  Well, it IS a problem, and here's why.  Autofocus sensors do not see color.  They only see black and white.  And when you desaturate this image, you get this:

When you take the color out of red, leaving only its gray value, it is nowhere near "middle gray."  In fact, as the histogram shows, it's barely half way between pure black and middle gray.  So your autofocus sensor actually has much less contrast to work with than our eyes think it should!

Now both these illustrations are for crisply defined edges.  What happens if the red object was, say, a soft fuzzy pillow?  Here's the same red-on-black image with the edges blurred:

Now there's even less contrast to work with.  Your autofocus would have a tough time with this image!

Sometimes autofocus can't handle things because they are too close together, or because the high-contrast region picked up by the sensor blocks out the subject you DO want to have in focus. No autofocus system could handle the image below.

At 100% crop (of a downsized jpg, so it looks lousy) you can see that the girl is actually in focus while the bamboo screen is not.  This HAD to be done manually, because autofocus would grab the screen and not the model.  I know; I tried it!

A couple of additional comments and we're done.  First a lot of red in any image tends to make a very poor black-and-white conversion.  Because red is actually such a dark color, red tones tend to convert as dark grey or near black.

That's a digression from talking about autofocus, but it's something worth remembering the next time you try to render a very red sunset as a black and white image!  (Yellow, on the other hand, converts splendidly, as you can see in the two sample pix below. The histogram shows you that the ONLY change made from the color version to black-and-white is desaturation.  Autofocus had no problem with this one.   Enjoy!)