Hands on: Reducing noise in pictures

This is not the first time I’ve covered dealing with digital noise, but it is the kind of problem that’s going to be around for a while, despite advances in sensor technology.

Camera manufacturers seem intent on packing more and more sensor sites on to smaller and smaller chips, resulting in higher resolution images but at the cost of increased signal-to-noise ratios.

In most circumstances, noise isn’t something to worry about. At low ISO settings, you’d have to look carefully to spot it using anything but a camera phone. But if, like me, you do a lot of shooting in low light and don’t like using flash, noise is something you have to learn to deal with.

The first step is to ensure you get the exposure right. In low light, you’re often shooting using the camera’s maximum aperture and trying to minimise subject or camera movement by selecting a shutter speed above 1/30th of a second. You’ll have set the camera ISO to its maximum sensitivity – 800 or even 1,600 ISO.

In these circumstances, it’s tempting to try to steal a stop of exposure by, say, shooting at 1/30th instead of the indicated exposure of 1/15th to avoid the possibility of camera shake. The problem arises when you come to adjust the image.

‘Stretching’ the histogram using a levels adjustment to redistribute all the pixels bunched on the left of the chart across the full tonal range, thereby lightening the image, solves one problem but causes another. Because relatively few bits are used to record shadow detail, stretching the histogram in this way brings out a lot of noise along with any shadow detail there might be.

Of course, there are situations when a photo with a lot of noise is better than no photo at all, or one with camera shake, which is a lot more difficult to deal with than noise. So given that noisy pictures are sometimes the only ones you can get, what can you do to improve them?

Noise abatement
As little as two or three years ago, if you wanted to get rid of – or, more realistically – reduce noise in your images, the only option was a plug-in or a utility specifically designed for the job. These days, most photo editors include noise reduction filters, though they don’t all work the same way and some are better than others.

Corel Paint Shop Pro 9 was one of the first applications to incorporate a noise filter and, though in the current version – Paint Shop Pro Photo XI – it is little changed, the Digital Camera Noise Removal filter is still a sophisticated and effective tool.

Like most of Paint Shop Pro’s filter dialogue boxes, it contains before and after preview thumbnails, in this case with the added functionality of setting and displaying sampling regions used to differentiate noisy pixels from genuine image data.

There are three sampling regions set by default. To set your own, drag in the left-hand thumbnail. Obviously, you need to define image areas where the noise is worst, but you also need to steer clear of areas with lots of image detail. Selecting regions of flat colour (where noise is usually most evident in any case) produces the best results.