demosaicing

Demosaicing, also known as debayering, is the process of reconstructing a full-color image from the incomplete color samples captured by an image sensor arranged with a color filter array. In most digital cameras the sensor uses a Bayer filter or similar mosaic in which each photosite records only one of the three color channels (red, green, or blue). Demosaicing interpolates the missing two channels for each pixel to produce a complete RGB image.

The Bayer pattern is the most common, arranged in 2x2 blocks containing red, green, green, blue (RGGB). Other arrangements exist (BGGR, GRBG, GBRG) as well as alternative mosaic patterns used by some sensors.

Algorithms range from simple to sophisticated. Early methods include bilinear or nearest-neighbor interpolation. More advanced approaches are edge-aware or directional, attempting to preserve detail and reduce artifacts such as zippering. Some techniques operate in the frequency domain or use probabilistic models; recent trends employ machine learning, including convolutional neural networks, trained on pairs of raw mosaicked and ground-truth images. In practice, demosaicing is often integrated with other steps in the image pipeline, such as denoising and color-space conversion.

Challenges include color artifacts like false colors, zippering along edges, cross-color artifacts in high-contrast or noisy regions, and error propagation in adjacent pixels. The choice of method involves trade-offs among color fidelity, detail preservation, noise amplification, and computation time.

Demosaicing is a standard step in processing RAW image data and is closely tied to subsequent white balance, gamma correction, and color-space conversion.