Understanding how colors are captured, transformed, and displayed is essential for keeping images consistent across screens.

Human Visual System

The eye converts light into electrical signals using three types of cones that are most sensitive to blue, green, and red wavelengths. These signals are then processed by the brain to produce the colors we perceive.

• Three cone classes (S, M, L) correspond roughly to blue, green, and red.
• Rods provide luminance information in low‑light situations but do not detect color.
• Brain combines cone responses into three channels: luminance, red‑green, and blue‑yellow.
• Human sensitivity peaks in the green region, influencing how colors are weighted.
• Understanding this physiology guides the design of efficient color encodings.

RGB Color Model

The RGB model maps the three cone responses directly to digital values for red, green, and blue channels.

• Each pixel stores three numeric values, typically 8‑bit per channel.
• Linear combination of the three channels recreates the perceived color.
• Negative values can appear in theoretical calculations, which real displays cannot emit.
• RGB is device‑dependent; the same triplet may look different on two screens.
• Many image formats, such as PNG and JPEG, are built around RGB data.

CIE XYZ Color Space

The International Commission on Illumination (CIE) introduced XYZ to provide a device‑independent representation without negative numbers.

• Transformation matrix converts RGB to XYZ, preserving perceived brightness in the Y channel.
• Y correlates with human perception of luminance, making it valuable for tone mapping.
• X and Z capture chromatic information independent of brightness.
• All perceivable colors map inside a defined 3‑D shape called the CIE gamut.
• XYZ serves as a common reference when building device‑specific profiles.

Device‑Specific Color Profiles

Each display has a unique way of emitting light, so color management systems translate XYZ values to the device's native RGB.

• Calibration involves measuring the emitted light for primary colors and white point.
• Measured XYZ coordinates define a transformation matrix back to the device's RGB.
• Profiles are stored in ICC files and applied by operating systems or browsers.
• Consistent appearance across devices relies on accurate profile creation.
• See related guidance on selecting proper models for visual tasks.

Perceptual Compression Strategies

Image formats compress data while keeping changes below the threshold of human perception.

• Lossless methods retain every pixel but offer limited size reduction.
• Lossy algorithms prioritize preserving luminance details, as the eye is more sensitive to brightness variations.
• Chroma subsampling reduces color resolution because the visual system tolerates lower chroma fidelity.
• Edge‑preserving filters keep high‑frequency details that aid recognition.
• Further reading on how AI influences image pipelines can be found in AI adoption in visual workflows.