How to Control Image Noise

Fixed pattern noise is associated with long exposures and high temperatures. It usually appears as ‘hot pixels’ which show up as white or coloured blocks.

Fixed pattern noise usually appears as scattered white and/or coloured pixels. It’s removed with long-exposure noise reduction by causing the camera to take a second exposure that records only    the pattern of dots and then mathematically subtracting this pattern from the image  data. This doubles the exposure time.

Random noise is also granular, often appearing as a ‘salt-and-pepper’ pattern with dark pixels in bright regions and bright pixels in dark regions. It usually occurs with high ISO settings.  Banding noise appears as stripes and is introduced by the camera when it ‘reads’ data from the image sensor. (Banding noise is rare with modern DSLRs.)

Random noise usually appears as granularity with dots of different colours. It’s not easy to remove without reducing image sharpness.

Noise not only changes with exposure settings and camera models, it can also vary within an image. Shadowed regions usually contain more noise than highlights. Cameras with large sensors that have big pixels can collect more light and are, therefore, less susceptible to image noise than small-sensor digicams with high pixel densities. This is the main reason DSLRs deliver better performance at high ISOs than compact cameras.  

Most sensors are designed to work best between ISO100 and ISO 200. However, when you use a modern DSLR camera, noise is seldom visible unless you shoot with ISO settings above 6400 ““ and even then, it may only be seen when the image is enlarged substantially.

Noise Reduction

Because all pixels collect some noise, all cameras perform some level of noise-reduction processing on the JPEG files they produce. Sometimes it’s applied automatically; at other times at the user’s discretion.

More sophisticated cameras provide separate processing settings for long exposures and high sensitivities. Newer cameras use newer technology and can deliver images with less noise at similar ISOs than earlier cameras could achieve.

[Right:] Modern DSLRs include powerful noise-reduction controls, enabling users to shoot at ISO 12800 (used for this photo) and produce virtually noise-free images.

Long-exposure noise reduction usually involves a technique known as ‘dark frame extraction’, which is applied automatically in many cameras for exposures longer than one second. The camera records the scene and then takes a second exposure of the same length with the shutter closed to capture the noise pattern produced by the sensor. The noise pattern is mathematically subtracted from the image, leaving it virtually noise-free.

While dark frame subtraction effectively removes most of the noise created during long exposures, high ISO noise-reduction processing tends to work by blurring the edges of noise patterns and this can reduce the sharpness of images. Photographers are faced with choosing whether sacrificing some real detail is acceptable if it allows more noise to be removed.

.

Dark frame subtraction is used to suppress noise in long exposures. But it can tie up the camera for long periods of time. A 35-minute exposure to record star trails as streaks locked the camera for 70 minutes to allow noise reduction processing.

Processing technologies work on two main types of noise. Luminance noise consists mostly of variations in brightness patterns. Chrominance (or colour) noise appears as coloured speckles. The aim is to identify the type of noise and apply the right amount of processing to minimise it. We tend to find colour noise more objectionable than luminance noise so most cameras apply more noise reduction to it. When images are edited, most software allows the user to control colour and luminance noise reduction separately.    

Using High ISO Settings

The effect of image noise depends on how images are viewed. We can often live with images that contain some visible noise, whereas images blurred by excessive noise-reduction processing become unattractive. Noise is usually more obvious when images are displayed on-screen than when they are printed.

Although noise will become more noticeable when images are enlarged and cropped, if noise- affected images are printed at large output sizes, the noise may not be noticeable when the prints are viewed from the appropriate distance. (Printing converts pixels into dots, blurring the edges between them.)

High temperatures can boost image noise so it pays to be cautious about using the top ISO values when the temperature is above about 20 degrees Celsius. Combining high ISO settings with long exposures at temperatures of 30 degrees Celsius and above is virtually certain to produce noise-affected photos.

[This article is an excerpt from  Digital SLR Pocket Guide 3rd Edn]