Debayering

Debayering, also known as demosaicing, is the process of reconstructing a full-color image from the incomplete color samples captured by a color filter array on a digital image sensor. In most cameras the sensor uses a Bayer filter mosaic, a 2x2 pattern of red, green, green, blue (RGGB). Each sensor pixel records only one color, so debayering estimates the other two color components for each pixel to produce a complete RGB image.

The simplest methods interpolate linearly from neighboring samples. Bilinear interpolation uses adjacent pixels, while bicubic interpolation can yield smoother results. More advanced algorithms are edge-aware and directional, using image gradients to preserve edges while estimating missing colors. Representative approaches include adaptive homogeneity-directed demosaicing, gradient-based methods, and learning-based techniques such as convolutional neural networks trained on paired data.

Demosaicing can introduce artifacts such as color zippering along high-contrast edges, false colors around fine patterns, and moiré. The choice of method involves trade-offs among computational cost, reconstruction quality, and memory usage, and it interacts with subsequent processing steps such as white balance and gamma correction. Demosaicing is typically performed on linear sensor data before gamma encoding.

Applications include digital cameras, smartphones, and video pipelines, as well as scientific imaging where faithful color reproduction is important. See also demosaicing, Bayer filter mosaic.