Noise is a problem in digital cameras similar to grain in film cameras: it appears as random coloured dots sprinkled around the image, or as uneven colouration of what should be smoothly-coloured areas.
Noise in a digital camera has several sources, including readout-circuit noise and pixel non-uniformity. However the dominant component is simply the random arrival of photons at the pixel grid of a sensor. This has the technical name shot noise (as in buckshot). Thus, noise is already present in the light striking the sensor, but stands out more clearly when dividing the incoming photons into smaller and smaller "bins." For this reason, sensors with small pixels display the most noise. The marketing impetus to increase camera megapixel counts inevitably results in smaller pixels, increasing the challenge of keeping noise at an acceptable level. This has forced sensor designers into new technical innovations: Pixel microlenses capture a greater fraction of the available light; and chips designed with back-side illumination (BSI) which reposition the necessary chip wiring beneath the light-receptive silicon layer.
Noise reduction software is often employed by digital cameras to reduce the appearance of noise in an image. Each new generation of camera incorporates a CPU with greater processing power, making it possible to smooth out the appearance of noise on the fly. This can sometimes lead to unpleasant, smeary or waxy looking artifacts if done too aggressively. A more natural "film-like" appearance can be achieved by employing weaker suppression of luminance noise, with stronger reduction in chroma noise (color speckles).