Noise is an unwanted visual artefact in digital photography and scanning, manifesting as random, irregularly distributed coloured or luminance varying pixels superimposed over the image, most prominently in shadow areas, dark tones, and regions of uniform colour or tone where the underlying image detail is smooth and the random variations introduced by noise are most easily distinguished from genuine image content. Digital noise is the electronic equivalent of grain in film photography, and while it shares some visual similarities with film grain, it differs in character - film grain has a more organic, randomly distributed quality that many photographers find aesthetically pleasing, while digital noise tends to have a harsher, more visually distracting appearance, particularly the coloured chromatic noise that appears as random red, green, and blue pixels in affected areas.
Digital noise in camera images arises from two principal sources, each producing a slightly different visual character. Luminance noise, also known as luma noise, consists of random variations in the brightness of individual pixels and produces a fine, granular texture somewhat analogous to film grain. Chromatic noise, also known as chroma or colour noise, consists of random variations in the colour of individual pixels, appearing as speckled patches of incorrect colour - typically red, green, and blue - scattered through affected areas of the image. Both types of noise are generated by the fundamental physical limitations of the image sensor and the electronic circuits that amplify and process the sensor's output signal.
The most significant factor influencing the amount of noise in a digital image is the ISO sensitivity setting used at the time of capture. At low ISO settings, the sensor's photosite voltage output is amplified relatively modestly, and the electronic noise inherent in the amplification circuit is small relative to the signal, producing images with low noise levels and clean shadow areas. As the ISO is increased to enable photography in lower light conditions, the amplification applied to the sensor output is increased proportionally, and the electronic noise in the amplification circuit is amplified along with the genuine image signal, becoming progressively more visible in the final image as the ISO setting rises. Very high ISO settings - particularly the extreme sensitivities above ISO 3200 or 6400 now offered by many digital cameras - produce images with significant noise that can substantially degrade image quality if not managed carefully.
In addition to ISO related amplification noise, digital cameras also produce thermal noise - a form of noise generated by the heat produced within the image sensor during long exposures. As the sensor operates over an extended period, individual photosites accumulate thermally generated electrons that are indistinguishable from those generated by actual photons of light, producing a fixed pattern of bright pixels and hot pixels that overlays the genuine image signal and becomes increasingly prominent as exposure duration increases. Many cameras apply long exposure noise reduction - a process in which a second dark frame exposure of the same duration is automatically captured immediately after the primary exposure and its fixed pattern noise signal is subtracted from the primary image - to reduce the visibility of thermal noise in long exposure photography.
Noise in scanned images arises from a different but related set of causes. When a scanner attempts to capture detail in the shadow areas of a film original or print, the light levels reaching the scanner's sensor from these dark areas are very low, and the electronic noise in the scanner's sensor and amplifier circuits becomes significant relative to the weak signal, producing a noisy rendering of shadow detail. Scanners with insufficient dynamic range - the range of tonal values from the darkest shadow to the brightest highlight that they can capture simultaneously - are particularly prone to shadow noise, as they lack the sensitivity and signal to noise performance needed to extract clean image information from the darkest tones of the original. Higher quality scanners with greater dynamic range and lower electronic noise can capture shadow detail more cleanly, producing scans with better tonal depth and less shadow noise.
Several approaches are used to manage and reduce noise in digital images. In camera noise reduction applies processing algorithms within the camera to reduce noise before the image is recorded to the memory card, either as part of the standard JPEG processing pipeline or as an optional long exposure noise reduction function. Post processing noise reduction using dedicated software tools - such as the noise reduction functions in Adobe Lightroom and Camera Raw, or specialist applications such as DxO PureRAW, Topaz DeNoise, and Noise Ninja - analyses the image and applies sophisticated algorithms to identify and smooth random pixel variations while preserving genuine image detail and edge sharpness as faithfully as possible.
Interestingly, certain image editing applications include a noise filter that deliberately adds a controlled amount of random pixel variation to a digital image, simulating the appearance of photographic grain for creative effect. This deliberate addition of synthetic noise or grain can be used to give a digital image a more organic, film like quality, to unify the texture of composited elements from different sources, or to add visual interest and texture to areas of flat, featureless tone that would otherwise appear unnaturally smooth and digital in character.