RAW Processing Quiz: Demosaicing and Tone Curves

Bayer sensors capture only one color channel per pixel, arranged in a mosaic pattern. Demosaicing algorithms examine surrounding pixels to estimate the two missing color values at each site, reconstructing a full-color image from the incomplete raw data.

Aggressive demosaicing algorithms recover fine spatial detail but risk interpreting color patterns in the mosaic as false color, especially at high-contrast edges. Softer algorithms suppress false color but may sacrifice some fine texture.

Every sensor has a baseline electronic signal even in complete darkness. Subtracting this offset before scaling raw values ensures that true black is represented as zero in the processed file rather than as a residual gray value.

In RAW processing, white balance is applied by multiplying the red, green and blue channels by different amounts so that a neutral object ends up with equal values. Because it is just a gain on each channel, it can be changed later without loss.

Highlight reconstruction uses the color channels that did not clip to estimate the one that did. If only one channel clipped slightly, there is plenty of real data to work from. Once all three are clipped, nothing is left to rebuild.

A camera profile maps each camera's unique spectral response to a target color space, compensating for differences in how individual sensors register reds, greens, and blues. It determines the overall color character of the rendered image independently of white balance.

Scene-referred linear data has perceptually flat midtones and compressed shadows. A tone curve redistributes this to match how human vision and displays perceive brightness, boosting midtone contrast and giving the image its normal visual appearance.

A linear RAW file has had little or no tone curve applied, so its values stay proportional to the light in the scene. It looks flat and dark until a curve is applied, but it gives maximum flexibility in editing.

Sharpening exaggerates fine contrast, and noise is fine contrast too. Applying it after demosaicing and noise reduction means it works on real detail instead of amplifying grain or mosaic artifacts.

Noise reduction smooths small pixel variations, and fine texture such as hair, skin and foliage is made of small variations too. When it is applied too strongly or too early, it blurs that detail along with the noise, leaving a waxy, plastic look.

Color noise, also called chroma noise, appears as random red, green and blue speckles or blotches, especially in dark areas at high ISO. Brightness noise, or luminance noise, looks more like grain.

A LUT maps specific input color or tonal values to predefined output values in a single computational step. It is used to apply complex color grades, creative looks, or output transformations efficiently without recalculating a series of adjustments.

Human vision is more sensitive to changes in dark tones than in bright ones. Gamma encoding redistributes digital values to allocate more steps to the shadows where the eye discriminates most, preventing visible banding in gradients.

Clarity and microcontrast tools apply local contrast enhancement by raising edge contrast at spatial frequencies between large-scale tonal transitions and fine detail. The result is a sense of crisp texture without the halos that global sharpening can produce.

Each camera's built-in JPEG engine applies its own color and contrast look. A camera matching profile copies that look inside a RAW editor, so RAW files start out looking like the camera's own previews instead of the editor's default rendering.