4.1 Introduction

Argument: Color is arguably the most important feature of an image. This is because human perception of the visual world is fundamentally based on color. People naturally see and interpret the world in terms of colors.

Human Perception vs. Computer Representation: While humans perceive color intuitively, representing and understanding color computationally is a complex task.
Infinite Colors: There are an infinite number of colors in the real world. Creating and representing all these colors accurately is a challenge.

Trichromatic Palette: Computers use a trichromatic palette to represent colors. This means they mix colors using three primary components.
3D Color Vector: Each color in a computer is represented as a three-dimensional vector: $(c_{1}, c_{2}, c_{3})$ . Each element $c_{i}$ represents the intensity or contribution of one of the three primary colors.
3D Color Space: The set of all possible color vectors created by the trichromatic palette forms a 3D color space.

Definition of color space/Model: Different ways of defining and representing those colors create different color spaces or color models.
A color space can be determined, based on how each of the trichromatic colors is defined.

Common Color Spaces: The PDF lists several commonly used color spaces:
- RGB: Red, Green, Blue. This is the most common color space, especially for display devices.
- LUV: A color space designed to be perceptually uniform (equal numerical differences correspond to roughly equal perceived color differences).
- HSV/HSL/HSI: Hue, Saturation, Value/Lightness/Intensity. These spaces are more intuitive for describing colors in terms similar to how humans perceive them (what color, how intense, how bright).
- YCrCb: Used in image and video compression. Separates luminance (Y’) from chrominance (Cr, Cb - red and blue color differences).
- Other Color Space: CIE XYZ, xyY.

Color Space:
- Definition: Models for representing the values of individual pixels. They define how a single color is represented (e.g., as an RGB triplet).
- Purpose: Provides a framework for defining and organizing colors.
Color Features:
- Definition: Characteristics computed from regions or the entire image to describe color patterns and statistics. They are used to compare and classify images based on their color content.
- Purpose: Enable analysis and understanding of the color distribution and arrangement within an image, not just individual pixel values. Used for tasks like image retrieval, classification, and segmentation.
Relationship:
- Color spaces provide the basic building blocks (pixel color representations).
- Color features are higher-level descriptors built upon those building blocks, characterizing the overall color properties of an image or its regions.

The PDF lists several examples of color features. These examples illustrate the types of information that color features capture:

Color Histogram: A representation of the distribution of colors in an image. It counts how many pixels of each color (or color range) are present.
Color Moments (CM): Statistical measures like mean, variance, and skewness of the color channels (e.g., the average red value, the spread of green values).
Color Coherence Vector (CCV): Distinguishes between pixels of the same color that are part of a large connected region (coherent) and those that are isolated (incoherent). Captures some spatial information.
Color Correlograms: Capture the spatial correlation of colors. They show how the probability of finding a certain color pair changes with distance.

Quartz 4