Happiness is a smooth histogram
John Lennon sang “Happiness is a warm gun” on the famous White Album and almost 30 years later, U2 used it in a few concerts, it seems. Reaching happiness remains a popular aim, clearly! For a photographer, happiness is not expressed by a warm gun—although a warm CF-card may indicate happiness to some extent, because that means a lot of data are written onto it in a short time. In other words: you’re shooting away.
But the real happiness comes when, afterwards, you can convert the foton imprints on the sensor into a large and richly textured print. To do that you need a photo-file allowing all the post-processing you need. In other words, one with as many pixels as your camera allows (although there is little magic in the ever higher numbers of megapixels for us amateurs), but also with as much colour information as possible. It’s a principle used in science too: keep as much of the data as long as possible, because once it is reduced, you cannot get back to the original.
The more nuances of red/green/blue are recorded in the photo-file, the better they can be adjusted to convey the precise picture that you want to show. Technically, this has to do with the colour space and colour depth. For colour space, don’t be fooled by nice sounding words like “vivid” or “natural”—they all work in the narrow sRGB space, well on my new Sony camera, anyway (read the small print in your camera’s manual to find out; it took a lot of searching for me!). A wider gamut is Adobe RGB, and ProPhoto RGB is even better, shows Wikipedia.
As for the nuances per colour, the usual choice is between 8-bits and 16-bits per channel (red/green/blue). It’s explained very well in the manual of The Gimp:
Color Depth is simply the number of bits used to represent a color (bits per pixel : bpp). There are 3 channels for a pixel (for Red, Green and Blue). GIMP can supprt 8 bits per channel, referred as eight-bit color. So, GIMP color depth is 8 * 3 = 24, which allows 256 * 256 * 256 = 16,777,216 possible colors (8 bits allow 256 colors).”
Obviously, 16 bits has double the amount of nuances and is therefore better for post-production; the Gimp cannot handle this, Photoshop can. Too bad, for otherwise The Gimp is a great open-source programme. (I’m looking forward to version 3: “version 3.0, allowing for the addition of long requested features like support for 16+ bits per channel and adjustment layers, to name a couple”
Back to the smooth histogram. If you fiddle around with 8-bit colours, you will see “holes” in the histogram once you try to increase contrast, for instance, which will lead to ugly “steps” in what should be smooth colour gradients. Blue skies are very prone to such effects—quite a nuisance to landscape photographers! See the histogram with the “pin-stripes” above. This will not happen (easily) with 16-bit data. In David Noton’s words in a column in May’s issue of magazine Practical Photography: “happiness is a smooth histogram. So, if you care about your pictures, shoot RAW”. The lesson is: take pictures in RAW format (no compression of data but all pixels’ information is available—and adjustable, unlike in TIFF), and process them with 16-bit colour depth.
The whole picture, below, is made from the 16-bit version. The detail above is in 8-bit. You don’t see the “steps” in the picture itself (shich could happen!), only in the histogram. So this is not the best bad example, but maybe the picture as a whole is a good example of a nice photo?