THE HIDDEN SCIENCE OF EVERY PHOTO

Your camerainvents color.

THE ART OF SEEING01 - 03
AN INTERACTIVE FIELD NOTESCROLL TO DISCOVER ↓ZENTRIX / 001

INSIDE THE CAMERA

Every site sees
one color.

The other two color components must be reconstructed.

MEASURED ESTIMATED FROM NEIGHBORS

A conceptual visualization. Explore the actual reconstruction below.

Go to the experiment
FILTER ARRAY / RGGB12 × 12 SITES
SELECTED SITEROW 06 / COL 06
Red50%ESTIMATED
Green50%ESTIMATED
Blue100%MEASURED

Linear-light values from an idealized grayscale checkerboard. Tile colors show filter types.

HOVER OR TAP A SITE USE ARROW KEYS TO EXPLORE

THE EXPERIMENT / 01INTERACTIVE STUDY

FIRST, LOOK CLOSER

This image has
no color.

01 / 03

So why does our virtual camera see green and magenta?

01 ORIGINAL INPUTGRAYSCALE
THE SCENEBLACK + WHITE
02 CAMERA OUTPUTRECONSTRUCTED
THE INTERPRETATION RGB

Change the pattern size or shift the sensor to see the colors change.

01 PX
FINECOARSE

iThe sensor recorded one color value at each location. The missing values were estimated using nearby samples. For fine patterns, those estimates can introduce colors that were not in the scene.

IDEALIZED SIMULATION · NOT PHONE RAW DATASENSOR ALIGNMENT: RG / GB
Go deeper in the reconstruction lab
THE RECONSTRUCTION LAB / 02AN INTERACTIVE STUDY

TAKE CONTROL

Become the camera.

The sensor has captured its measurements. Now you decide how to reconstruct the missing colors.

ORIGINALRECONSTRUCTED
RECONSTRUCTION METHOD

Estimate missing colors by averaging nearby measurements. Averages nearby samples, which can soften edges and mix fine detail.

FALSE COLOR SCORE100.00Channel disagreement · grayscale only
RECONSTRUCTION ERROR0.5000Mean squared error · linear RGB
DETAIL RATIO0.00×Luminance gradient vs. original

Zero false color means the result remains gray. Error and detail are measured against the unfiltered original; no single metric tells the whole story.

THE CHALLENGE

Can you fix what
the camera invented?

The original pattern is black and white. Your reconstruction added color. Can you remove the false colors without losing important detail?

ORIGINALRECONSTRUCTED

FALSE COLOR 100.00 · ERROR 0.5000 · DETAIL 0.00×

RECONSTRUCTION METHOD

Estimate missing colors by averaging nearby measurements. Averages nearby samples, which can soften edges and mix fine detail.

Some high-frequency detail must be suppressed before sampling to avoid misleading measurements. This Gaussian blur is an approximation, not a complete optical model.

Try it with a photograph
YOUR PHOTO / 03YOUR IMAGE, YOUR DEVICE

LOOK CLOSER

What did your camera
never see?

Upload a photograph and explore how a simulated Bayer sensor could reconstruct its colors.

Sample · processed in your browser

This experiment simulates Bayer capture from an existing image. It does not recover your camera's original sensor data. Images stay on your device.

THE PROCESS / 02FROM LIGHT TO IMAGE

A LITTLE SENSOR SCIENCE

One measurement.
Three colors.

Every photograph begins with an incomplete picture. A remarkably clever process fills in the rest.

01 / 03↗

Measure

A filter lets each sensor site record one color component.

02 / 03↗

Estimate

Neighboring samples provide clues about what is missing.

03 / 03↗

Reconstruct

The processor combines measured and estimated values into a full-color image.

MORE GREEN SITES HELP CAPTURE DETAILR G / G B
A NEW WAY OF SEEING

You've seen the image.
Now see the invisible.

The next time you take a photograph, remember: part of its color was calculated.

Experiment again