Camera Color

A digital sensor is a grid of light buckets. Each photosite counts photons and reports a single number - how much light, not what color. On its own a sensor would only ever produce a grayscale image. To capture color, the camera puts a color filter over each pixel so that pixel responds to only red, only green, or only blue light. Color, in a camera, is reconstructed from a mosaic of single-color measurements - not measured directly.

That single design choice - filter, then reconstruct - ripples through everything: why files need demosaicing, why white balance is a set of gains, why a color matrix exists, and why raw preserves so much more than JPEG. The rest of the article follows the data from photosite to finished pixel.

The dominant filter pattern is the Bayer array: a repeating 2×2 tile of one red, two green, one blue (RGGB). There are twice as many green sites because our eyes are most sensitive to green and it carries most of the luminance detail. The sensor's raw output is therefore a mosaic - each pixel holds just one channel. Demosaicing reconstructs the full image by interpolating the two missing channels at every pixel from its neighbors.

The raw sensor data is not white-balanced - under warm light the red channel reads high and blue low, and vice versa. White balance in a camera is simply a set of per-channel gains (usually scaling red and blue relative to green) that make a neutral object read equal across channels. Because raw stores the data before these gains are baked in, you can set white balance freely after the fact with no quality loss.

Even white-balanced, the sensor's raw RGB is not "real" color - the filter sensitivities do not match human cones, so colors come out dull and slightly off, with channels that overlap differently than our eyes. A color matrix (a 3×3 transform, part of the camera's color profile) maps the camera's native response into a standard space like sRGB or a profile connection space. This is the step that turns flat sensor data into accurate, saturated color - and why two cameras need different profiles.

A JPEG is a finished photo: the camera has already demosaiced, white-balanced, applied the color matrix and tone curve, sharpened, and compressed to 8 bits - all decisions baked in. A raw file stores the high-bit sensor data with those steps recorded as editable instructions. The biggest practical payoff is highlight recovery: a raw file often holds detail in bright areas that look pure white in the JPEG, because the JPEG has already clipped them to 255.

Putting it together, light becomes a finished color image through a fixed chain - whether the camera does it instantly for a JPEG or you do it later from raw.

Six questions on the Bayer array, demosaicing, white balance, and raw. Instant feedback, no scores recorded - a wrong answer comes with a short explanation.

Capture is the front of the pipeline whose later stages this library covers. The numbers reflect each article's position in the editorial roadmap.