Speckling is a digital image artefact characterised by the appearance of isolated, randomly distributed bright or light coloured pixels scattered through predominantly dark or shadow areas of an image, creating a visually distracting spotted or speckled texture that disrupts the smooth, continuous tone that should characterise these dark regions. The individual bright pixels that constitute speckling appear as tiny, isolated points of lighter tone surrounded by correctly rendered dark pixels, giving the affected areas a granular, noisy quality quite different from the smooth, clean dark tones that the image should ideally display.
Speckling most commonly arises in two distinct contexts - scanning and digital camera capture - with somewhat different underlying causes in each case. In scanning applications, speckling in dark areas of the scanned image is typically caused by the electronic noise inherent in the scanning device's sensor and amplification circuitry, which becomes significant relative to the very weak signal produced by the small amounts of light reflected or transmitted from dark areas of the original being scanned. When a scanner attempts to capture detail in the deepest shadow areas of a film original or reflective print, the light levels reaching its sensor are extremely low, and the random electronic noise in the sensor and amplifier - which is present at all times regardless of the light level - becomes a larger proportion of the total signal, causing individual sensor elements to produce incorrect, randomly elevated readings that appear as isolated bright pixels in the dark areas of the scanned image.
Scanners with insufficient dynamic range - the range from the darkest shadow to the brightest highlight that they can capture with acceptable signal to noise performance - are particularly prone to producing speckling in shadow areas, as the limited sensitivity and dynamic range of their sensors mean that the signal from very dark areas falls below the level at which it can be distinguished reliably from the background noise of the electronic system. Higher quality scanners with wider dynamic range and lower electronic noise floors can capture dark area information more cleanly, producing scans with less speckling and better shadow detail at the extremes of the tonal range.
In digital camera images, speckling in dark areas is closely related to the phenomenon of digital noise, arising from thermal noise in the image sensor - where heat generated by the sensor during operation causes individual photosites to produce spurious electron counts that register as bright pixels - and from the fixed pattern noise and random read noise introduced by the sensor's readout circuitry. At high ISO settings, where the signal from the sensor is amplified to compensate for low light levels, the random noise in the amplification circuit is amplified along with the genuine image signal, producing the pronounced speckling in shadow areas that characterises high ISO digital images. Long exposure photography similarly produces speckling through the accumulation of thermal noise in the sensor over extended exposure periods.
Speckling is distinct from the regular, structured artefacts produced by specific types of sensor malfunction - such as hot pixels, which are consistently bright pixels appearing at fixed locations in every image due to permanently defective photosites - in that speckling is random in distribution and varies between frames, appearing in different locations in different images rather than always occurring at the same fixed positions. This random character reflects the underlying randomness of the noise processes that generate speckling and distinguishes it from systematic sensor defects that require different remediation approaches.
The practical remediation of speckling in digital images is typically addressed through noise reduction processing, either applied within the camera as part of the in-camera JPEG processing pipeline, or applied in post-processing software using dedicated noise reduction tools. Noise reduction algorithms analyse the spatial distribution of pixel values in the image and identify and smooth the random variations that constitute noise while attempting to preserve the genuine, structured image detail that should be retained. In scanning workflows, speckling can be reduced by using a scanner with adequate dynamic range for the tonal range of the original being scanned, scanning at higher bit depths that provide greater tonal precision in shadow areas, and using multiple scan averaging - making several scans of the same original and averaging the pixel values between them - to reduce the random noise component through statistical averaging.