White Balance is Broken
Based on minutephysics's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.
Equal Kelvin steps in white balance do not correspond to equal color changes because the temperature-to-color relationship is nonlinear (Wien’s law).
Briefing
Professional cameras and editing tools often make white-balance adjustments in equal steps of Kelvin, but equal Kelvin steps do not translate into equal changes in the actual color of light. The result is a practical mismatch: small Kelvin tweaks at the low end of the temperature range can swing images dramatically toward blue or yellow, while the same-sized Kelvin change at the high end produces a comparatively subtle color shift. That uneven “color spacing” helps explain why many real-world photos and videos end up with stubborn, hard-to-correct color casts.
The core issue comes from physics. Light from glowing hot objects follows Wien’s law, which links color to temperature through a nonlinear relationship (roughly an inverse dependence on temperature). Yet most digital white-balance controls behave as if the mapping from Kelvin to perceived color is linear. A sequence of images spaced by equal Kelvin increments illustrates the imbalance: the low-temperature side yields only a few blue-leaning frames, while the high-temperature side produces many yellow-leaning frames. When the same images are instead spaced by equal color increments, the progression looks more even—suggesting that “equal color change” is the quantity that should be stepped through, not equal Kelvin.
This nonlinearity also affects how users interpret the scale. At low Kelvin values, a small adjustment (for example, moving from 2700 to 2800 K) can cause a large color difference, making it easy to overcorrect and end up too blue or too yellow. At higher Kelvin values, the same numerical step corresponds to a smaller color shift, so mistakes are less visually dramatic. Even so, the UI can mislead operators into thinking the color changes are uniform, which can lead to undercompensation for the lighting.
The transcript draws a parallel to exposure controls, where cameras typically use increments designed for human perception. Human brightness perception is logarithmic: doubling light doesn’t feel like a huge jump, so exposure settings are offered in “equal perceptual steps” (commonly described as stops). Cameras and software handle exposure this way, but white balance is still commonly offered as linear Kelvin steps—despite being nonlinear in the underlying physics.
Historically, film-era workflows used color-correction filters and gels that were effectively spaced by equal color change, not equal Kelvin numbers. The shift to digital seems to have preserved the convenient “round number” Kelvin stepping without carrying over the correct nonlinearity. The transcript speculates that this mismatch may stem from the way white balance was implemented for digital systems rather than from any deliberate photographic design.
The practical takeaway is twofold. Photographers and videographers should be especially careful when dialing white balance in the low Kelvin range, where small numeric changes have outsized color impact. For camera makers and software companies, the proposed fix is straightforward: keep Kelvin as the familiar label if desired, but offer increments that correspond to equal color steps (analogous to how exposure uses stops). Canon’s cinema cameras are cited as already moving toward this approach by allowing equal color increments, while many other systems—including those using equal Kelvin increments—remain vulnerable to the same “broken spacing” problem.
Cornell Notes
Equal Kelvin steps in white balance do not produce equal color changes. That’s because the color of light from hot objects follows Wien’s law, a nonlinear relationship between temperature and emitted spectrum, while most digital cameras and editing software treat Kelvin as if it were linear. The mismatch makes low-Kelvin adjustments (e.g., around 2700 vs 2800 K) far more sensitive, while high-Kelvin adjustments (e.g., 5000 to 5350 K) change color less. The transcript argues that white balance should be controlled in increments that correspond to equal color “stops,” similar to how exposure uses logarithmic stops. Canon’s cinema cameras are presented as an example of closer-to-correct behavior via equal color increments.
Why do equal Kelvin increments produce uneven color changes?
How does this nonlinearity affect real-world white-balance mistakes?
What comparison does the transcript make to exposure controls?
What would a better white-balance control scheme look like?
Is RAW a complete fix for white-balance problems?
Review Questions
- If a camera offers white balance in equal Kelvin increments, what physical law makes that stepping inherently non-uniform in color change?
- Why are white-balance adjustments near 2700 K described as more risky than similar numeric adjustments near 9000–10000 K?
- What design principle used for exposure (stops/logarithmic perception) does the transcript argue should be applied to white balance?
Key Points
- 1
Equal Kelvin steps in white balance do not correspond to equal color changes because the temperature-to-color relationship is nonlinear (Wien’s law).
- 2
Low Kelvin adjustments are more sensitive: small numeric changes can cause large blue/yellow shifts, increasing the chance of overcorrection.
- 3
High Kelvin adjustments are less sensitive: the same Kelvin step produces a smaller visible color change, so mistakes are less dramatic.
- 4
Most cameras and editing tools treat white balance as linear in Kelvin even though exposure controls are designed around nonlinear perception (stops).
- 5
A better approach is to offer white-balance increments based on equal color change (“stops”), not equal Kelvin differences, while optionally keeping Kelvin as the label.
- 6
Canon’s cinema cameras are cited as closer to equal color increments, while many other systems still rely on equal Kelvin stepping.
- 7
RAW can help after the fact, but it’s not always available (especially for video) and doesn’t address the underlying control mismatch.