How does a Bayer filter capture color and what is demosaicing?

WHAT THIS TESTS: This question probes whether you understand the difference between physical light capture and computational reconstruction in digital imaging. A senior candidate should know that a single sensor cannot simultaneously record full spectral information at every spatial location, so the system relies on a color filter array and post-processing interpolation. The interviewer is looking for awareness of spatial-spectral tradeoffs, the reason green is sampled more frequently than red or blue, and the algorithmic complexity of reconstructing a full RGB image from sparse samples.

A GOOD ANSWER COVERS: First, describe the Bayer color filter array as a repeating grid of red, green, and blue filters overlaid on the sensor, usually in an RGGB pattern where half the pixels are green and a quarter each are red and blue. Second, explain that each photosite measures only one color channel, so the raw image contains incomplete color information. Third, define demosaicing as the interpolation step that estimates the two missing color values for every pixel by examining neighboring values, and note that high-quality algorithms use edge-directed or adaptive methods rather than naive averaging to avoid blurring fine detail. Fourth, name a concrete artifact such as moire patterns in repetitive textures, zippering along sharp edges, or false color fringing near high-contrast boundaries.

COMMON WRONG ANSWERS: A major red flag is claiming that every pixel natively captures full RGB data, which would require a three-sensor beam-splitter architecture or a Foveon-style stacked sensor, not a standard Bayer design. Another weak response is describing demosaicing as simple bilinear averaging across neighbors without acknowledging that edge direction and local structure matter. Some candidates also confuse demosaicing with debayering terminology or conflate it with white balance and gamma correction, revealing fuzzy mental models of the image processing pipeline.

LIKELY FOLLOW-UPS: An interviewer might ask why green is prioritized in the Bayer pattern, which connects to human luminance sensitivity and the need for higher spatial resolution in the channel that carries most brightness information. They might also ask how to reduce demosaicing artifacts, leading to discussion of optical low-pass filters, oversampling with higher-resolution sensors, or computational photography approaches like multi-frame super-resolution. A senior candidate could be pressed on alternative CFAs such as X-Trans or quad-Bayer, or on whether machine learning has replaced traditional interpolation kernels.

ONE CONCRETE EXAMPLE: Imagine photographing a fine checkered shirt with a 24 megapixel Bayer sensor. The red and blue pixels are too sparse to resolve the tight pattern, so the demosaicing algorithm guesses edge directions between green samples. If it guesses wrong, the shirt develops a swirling moire pattern or rainbow-like false colors because the interpolation aliases high-frequency luminance into chroma channels. A better algorithm or an optical low-pass filter that slightly blurs the image before sampling can suppress this artifact at the cost of some sharpness.