Digital Image Processing and Color Perception

Digital Image Processing (DIP)

• Digital Image Processing (DIP) is a field studying image manipulation.
• A website for learning about ImProc is available at https://moodle.eurecom.fr/course/view.php?id=415
• An introduction to Colorimetry was presented on November 13, 2024.
• The website confirmed that the dress is blue and black.

Color Illusions

• Dress images caused an internet discussion about color perception.
• The dress was initially posted on Tumblr.
• The original image caused many people to interpret the colors as white and gold, and some as blue and black.
• The perceived color difference is due to lighting conditions and differences in individual color perception.

Color Perception

• Color perception depends on the source of illumination.
• The characteristics of the object (absorptions, wavelengths).
• Characteristics of the human eye (example: cones).
• The human brain's color interpretation.

Light

• Light is electromagnetic radiation.
• The human eye registers a limited spectrum of electromagnetic radiation (between 780nm and 380nm, which is less than 1% of the total emitted by the sun).
• Wavelength determines color and amplitude determines brightness.

Light Sources

• Light comes from various sources, and the light source affects how colors are perceived.
• Common light sources include incandescent light (sunlight, candles, light bulbs), gas discharge (sodium, mercury, xenon lamps), photoluminescence (fluorescent lighting), and chemical reactions.

Illuminants

• Light sources, or illuminants, are defined by color temperature and spectral power distribution.
• Different light sources have different spectral power distributions.
• Color temperature is a measure of the color of a light source.

Human Eye

• The retina is the membrane in the eyeball that receives images from the lens, which changes the image into nerve impulses.
• The fovea is the central part of the retina, with high visual acuity and many cone cells.

Photosensitive Cells

• Two types of photosensitive cells are in the retina.
• Rods are sensitive to brightness and luminance (scotopic vision). The number of rods is approximately 120 million.
• Cones are sensitive to color and frequency (photopic vision). The number of cones is approximately 5 million.
• Cones are more concentrated in the fovea.
• Rods predominantly occur in the periphery of vision.

Cone Facts

• Color vision comes from three types of cones, each responding to different colors; red (L), green (M), and blue (S).
• The sensitivity peaks in the green range.
• Red cones account for 64% of overall absorption, green for 32%, and blue for 4% (although blue cone density is not present in the center of the retina).
• The perception of a spectrum of colors is produced by stimulating these cones in different ways.
• Metamers differ by light spectrum but produce the same color perception.

Color Illusions

• Color illusions involve visual perception that differs from reality.
• Lightness contrast: surrounding dark colors make parts of the image appear lighter than they are, and vice-versa.
• Saturation contrast: a strong color stands out more against a neutral background.
• Hue contrast is similar to lightness contrast.
• Complementary contrast is described when a color appears much brighter against its complementary color.

Color Blindness

• Color blindness is a deficiency in color vision.
• Various types of color blindness exist.
• Anomalous Trichromacy is a condition of color perception where one of the cones has lost some ability to distinguish colors (6.3% of males, 0.37% of females).
• Protanomaly is a type of anomalous Trichromacy.
• Deuteranomaly is a type of anomalous Trichromacy.
• Tritanomaly is a type of anomalous Trichromacy.
• Dichromacy is a general term for a color vision defect where one type of cone is missing.
• Protanopia has missing L-cones (1.3% of males, 0.02% of females).
• Deuteranopia has missing M-cones (1.2% of males, 0.01% of females).
• Tritanopia has missing S-cones (0.001% of males, 0.03% of females).
• Rod monochromacy has missing cones, and only rods (0.00001% of males, 0.00001% of females).

Ishihara Test

• The Ishihara test is used to detect color blindness.
• The test includes figures with various colors and patterns.

Color Naming

• Color naming is attaching labels like black or red to colors.
• Perceptual categories are constructed through language.
• Culture colors (Berlin & Kay, 1969): black, white, red, green, yellow, blue, brown, purple, pink, orange, gray.

Additive Color Mixing

• Primary colors: red, green, and blue.
• Secondary colors: magenta, yellow, and cyan.
• Starting with a black background, adding colored lights produces color.
• Used in television, cameras, computer graphics, etc.

Additive Color Mixing Rules

• Grassman's laws describe how colors are combined.

  • Law 1: any color can be matched by combining primary colors linearly.
  • Law 2: a mixture of two colors can be matched by linearly adding their combined components.
  • Law 3: proportionality.

Subtractive Color Mixing

• Primary colors: yellow, cyan, magenta.
• Secondary colors: red, green, and blue.
• Colors are created by subtracting colors from white.
• Used in printing and photography.

White Balance, Bayer Interpolation

• A white object will have equal reflectance values for each primary color (R=G=B).
• Bayer interpolation is a method to create a color image using a monochrome image sensor.
• Each sensor, in the Bayer array, only measures one color value within the image.
• R or B value interpolations are conducted from neighboring pixels of the same color.

Color Models

• Color models are systems to categorize and define colors based on attributes like hue, saturation/chroma, and lightness/value.
• Examples include: Munsell, Chevreul, Ostwald, RGB, HSV, CIE XYZ, CIE Lab, and CIE Luv.

RGB Color Model

• RGB is a color model that uses red, green, and blue light intensities to reproduce color.
• The color space can be represented as a cube.

Problems with RGB

• The perceived range of colors in RGB is small.
• Calculating the correct combination of colors is difficult.
• The model is not perceptually linear (distance apart does not necessarily imply a similar difference in perception.)

Color Making Attributes

• Hue: A color's distinctive quality, like red, orange, or blue.
• Lightness/Value: The quantity of light reflected by a color, influencing how light or dark it appears.
• Chroma/Saturation: A color's intensity or purity; a highly saturated color is bright while a less saturated color is dull.

Munsell Color System

• Munsell created a system for describing colors using decimal notation.
• The system, modeled as an orb, uses a central column representing the value of colors, each arm representing color hue, and each segment's length denoting saturation.

HSV/HLS Color Models

• HSV (Hue, Saturation, Value): A color model that describes colors in a simpler, more intuitive way for designers/artists. The model describes colored lights in comparison to their respective intensities, rather than mixing multiple colored lights.
• HLS (Hue, Lightness, Saturation): A similar model to HSV, based on brightness more than color intensity.

CIE XYZ Color Model

• Defined in 1931, described the full space of perceptible color.
• Uses hypothetical primary colors to replace red, green, and blue.
• Mathematical representation of colors, but not physically attainable.

Color Matching

• Some colors aren't achievable by simply combining primary colors.
• A solution is to add light from the other side.

CIE xyY Color Space

• uses x, y, and Y for representing chromaticity coordinates and brightness.

Dominant Wavelength

• Describes the color of light that will form a given color.
• The purity (saturation) of a color is the ratio of distance between a given point in color space (that is, the color in question).

Complementary Wavelength

• Complimentary colors are colors that produce white light when mixed.
• Colors located opposite each other in a color model (like a circle) are complementary.
• Colors that create white light when mixed.

Perceptual Uniform Color Model

• Distance between colors in the model closely matches perceived distances.

Non-perceptually Uniform Color Models

• Colors in some models are not perceived to have a distance in color space that matches the Euclidean distance. (a measure of distance often calculated from points on a graph)

CIE Lab Color Space

• Developed in 1976.
• A perceptually uniform color space, meaning distance in the model correlates with the perceived distance between colors.
• Uses three variables: L (luminance), a (red-green axis), and b (blue-yellow axis).

YIQ Color Model

• Used in US commercial color TV broadcasting.
• Y encodes luminance (brightness).
• I and Q encode color differences.
• YIQ is more used for broadcast signals due to signal bandwidth needs.

YCrCb Color Model

• For analog video, now used in digital video.
• Y represents luminance.
• Cr and Cb represent color differences.

True Color Image

• A true color image uses 24 bits to represent a color.

Look-Up Table

• A table that maps pixel values (numbers) to colors.
• Used to define colors for images with fewer colors.

Color FAQ

• A common resource listing questions and answers about color.

Anaglyphs

• Images using red-cyan filters to give the appearance of stereo vision.

Explore the fascinating world of Digital Image Processing (DIP) and how it relates to color perception. This quiz delves into concepts like colorimetry and the famous dress illusion that sparked debates on color interpretation. Test your knowledge on how lighting and human biology affect our perception of colors.