Lecture Outline

Color perception and representation

Human perception of color Trichromatic color mixing theory Different color representations

Lecture 8. Color Image Processing

EL512 Image Processing Dr. Zhu Liu zliu@research.att.com
Note: Part of the materials in the slides are from Gonzalezs Digital Image Processing and Onurs lecture slides

Color image display

True color image Indexed color images

Pseudo color images Color image enhancement

Fall 2004 EL512 Image Processing Lecture 8, Page 2

Light is part of the EM wave

Human Perception of Color

Retina contains receptors
Cones
Day vision, can perceive color tone Red, green, and blue cones

Rods
Night vision, perceive brightness only

Color sensation
Luminance (brightness) Chrominance
Hue (color tone) Saturation (color purity)
Fall 2004 EL512 Image Processing Lecture 8, Page 3 Fall 2004 EL512 Image Processing

From http://www.macula.org/anatomy /retinaframe.html

Lecture 8, Page 4

Frequency Responses of Cones

Illuminating and Reflecting Light

Illuminating sources:
emit light (e.g. the sun, light bulb, TV monitors) perceived color depends on the emitted frequency

Reflecting sources:
reflect an incoming light (e.g. the color dye, matte surface, cloth) perceived color depends on reflected frequency (=emitted frequency - absorbed frequency

Ci = C ( )ai ( )d , i = r , g , b, y
Fall 2004 EL512 Image Processing Lecture 8, Page 5 Fall 2004 EL512 Image Processing Lecture 8, Page 6

Tri-chromatic Color Mixing

Tri-chromatic color mixing theory
Any color can be obtained by mixing three primary colors with a right proportion

RGB vs CMY

Primary colors for illuminating sources:

Red, Green, Blue (RGB) Color monitor works by exciting red, green, blue phosphors using separate electronic guns follows additive rule: R+G+B=White

Primary colors for reflecting sources (also known as secondary colors):

Cyan, Magenta, Yellow (CMY) Color printer works by using cyan, magenta, yellow and black (CMYK) dyes follows subtractive rule: R+G+B=Black
Fall 2004 EL512 Image Processing Lecture 8, Page 7

Magenta = Red + Blue Cyan = Blue + Green Yellow = Green + Red

Fall 2004

Magenta = White - Green Cyan = White - Red Yellow = White - Blue

EL512 Image Processing Lecture 8, Page 8

A Color Image

Tristimuls Values
Tristimulus value
Red

The amounts of red, green, and blue needed to form any particular color are called the tristimulus values, denoted by X, Y, and Z. Trichromatic coefficients
x= X , X +Y + Z y= Y Z , z= . X +Y + Z X +Y + Z

Green

Blue

Only two chromaticity coefficients are necessary to specify the chrominance of a light.

x + y + z =1

Fall 2004

EL512 Image Processing

Lecture 8, Page 9

Fall 2004

EL512 Image Processing

Lecture 8, Page 10

CIE Chromaticity Diagram

CIE (Commission Internationale de LEclairage, International Commission on Illumination ) system of color specification
x axis: red y axis: green The point marked with GREEN x: 25%, y: 62%, z: 13%.

Color Models
Specify three primary or secondary colors
Red, Green, Blue. Cyan, Magenta, Yellow.

Specify the luminance and chrominance

HSB or HSI (Hue, saturation, and brightness or intensity) YIQ (used in NTSC color TV) YCbCr (used in digital color TV)

Amplitude specification:
8 bits per color component, or 24 bits per pixel Total of 16 million colors A 1kx1k true RGB color requires 3 MB memory
Lecture 8, Page 11 Fall 2004 EL512 Image Processing Lecture 8, Page 12

Fall 2004

EL512 Image Processing

RGB Color Model

CMY and CMYK Color Models

Conversion between RGB and CMY
C 1 R M = 1 G , Y 1 B R 1 C G = 1 M . B 1 Y

RGB 24-bit color cube

Equal amounts of Cyan, Magenta, and Yellow produce black. In practice, this produce muddy-looking black. To produce true black, a fourth color, black is added, which is CMYK color model.
Fall 2004 EL512 Image Processing Lecture 8, Page 14

Fall 2004

EL512 Image Processing

Lecture 8, Page 13

HSI Color Model

Hue represents dominant color as perceived by an observer. It is an attribute associated with the dominant wavelength. Saturation refers to the relative purity or the amount of white light mixed with a hue. The pure spectrum colors are fully saturated. Pink and lavender are less saturated. Intensity reflects the brightness.
Fall 2004 EL512 Image Processing Lecture 8, Page 15 Fall 2004

The HSI Color Model

EL512 Image Processing

Lecture 8, Page 16

Conversion Between RGB and HSI

Converting color from RGB to HSI
H = 360 S = 1 , with = cos 1 1 if B > G 2 ( R G ) + ( R B)(G B) 2 if B G 1 [( R G ) + ( R B)] 2

YIQ Color Coordinate System

YIQ is defined by the National Television System Committee (NTSC)
Y describes the luminance, I and Q describes the chrominance. A more compact representation of the color. YUV plays similar role in PAL and SECAM.

3 [min( R, G , B)] ( R + G + B)

1 I = [ R + G + B] 3

Converting color from HSI to RGB

RG sector (0H<120)
B = I (1 S ) S cos H R = I 1 + cos(60 H ) G = 1 ( R + B)
Fall 2004

GB sector (120H<240)
R = I (1 S ) S cos( H 120) G = I 1 + cos(60 ( H 120)) B = 1 (R + G)
EL512 Image Processing

BR sector (240H<360)
G = I (1 S ) S cos( H 240) B = I 1 + cos(60 ( H 240)) R = 1 (G + B )
Lecture 8, Page 17

Conversion between RGB and YIQ

0.114 R Y 0.299 0.587 I = 0.596 0.274 0.322 G , Q 0.211 0.523 0.311 B
Fall 2004

0.621 Y R 1.0 0.956 G = 1.0 0.272 0.649 I B 1.0 1.106 1.703 Q

Lecture 8, Page 18

EL512 Image Processing

Criteria for Choosing the Color Coordinates The type of representation depends on the applications at hand.
For display or printing, choose primary colors so that more colors can be produced. E.g. RGB for displaying and CMY for printing. For analytical analysis of color differences, the difference in the trisitumulus values are linearly related to the chrominance differences. HSI is more suitable. For transmission or storage, choose a less redundant representation, eg. YIQ or YUV
Fall 2004 EL512 Image Processing Lecture 8, Page 19

Comparison of Different Color Spaces

Fall 2004

EL512 Image Processing

Lecture 8, Page 20

Demo Using Photoshop

Color Image Display and Printing

Display:
Need three light sources projecting red, green, blue components respectively at every pixel Analog display: raster scan Digital display: directly projecting at all pixel locations

Printing:
Need three (or more) color dyes (Cyan, Magenta, Yellow, and Black) Analog printing Digital printing Out of gamut color
Fall 2004 EL512 Image Processing Lecture 8, Page 21 Fall 2004 EL512 Image Processing Lecture 8, Page 22

Color Image Display

Input 0 1 255 Red signal Red LUT Green LUT Blue LUT Red Gun Red Gun Red Gun Output

Color Gamut

Green signal

Blue signal

Each color model has different color range (or gamut). RGB model has a larger gamut than CMY. Therefore, some color that appears on a screen may not be printable and is replaced by the closest color in the CMY gamut. Lecture 8, Page 23 Fall 2004 EL512 Image Processing Lecture 8, Page 24

Fall 2004

EL512 Image Processing

Demo with Photoshop

Using photoshop to show how out of gamut color (view->Gamut warning).

Gamma Correction
The intensity to voltage response curve of the computer monitor is not linear.
Sample Input to Monitor Output from Monitor Graph of Input Graph of Output L=V2.5

Gamma correction
Sample Input to Monitor Graph of Input Graph of Correction L=L1/2.5 Graph of Output
EL512 Image Processing Lecture 8, Page 26

Gamma corrected Input Monitor Output

Fall 2004 EL512 Image Processing Lecture 8, Page 25 Fall 2004

Color Quantization
Select a set of colors that are most frequently used in an image, save them in a look-up table (also known as color map or color palette) Any color is quantized to one of the indexed colors Only needs to save the index as the image pixel value and in the display buffer Typically: k=8, m=8 (selecting 256 out of 16 million)
Input index (k bits) Index 1 . Index 2^k
Fall 2004

Uniform vs. Adaptive Quantization

Uniform (scalar quantization)
Quantize each color component uniformly
E.g. 24 bit-> 8 bit can be realized by using 3 bits (8 levels) for red, 3 bits (8 levels) for green, 2 bits (4 levels) for blue Do not produce good result

Adaptive (vector quantization)

Treating each color (a 3-D vector) as one entity. Finds the N colors (vectors) that appear most often in a given image, save them in the color palette (codebook). Replace the color at each pixel by the

Red color (m bits)

Green color (m bits)

Blue color (m bits)

Lecture 8, Page 27

closest color in the codebook

The codebook (I.e. color palette) varies from image to image -> adaptive
Fall 2004 EL512 Image Processing Lecture 8, Page 28

EL512 Image Processing

Illustration of the Vector Quantization

y Codebook size: 25 y

Example of Color Image Quantization

24 bits -> 8 bits

Uniform Quantization
Fall 2004 EL512 Image Processing

Vector Quantization
Lecture 8, Page 29

Adaptive (non-uniform) quantization (vector quantization)

Fall 2004

Uniform quantization (3 bits for R,G, 2 bits for B)

Lecture 8, Page 30

EL512 Image Processing

Web Colors: 216 Safe RGB Colors

Color Dithering
Color quantization may cause contour effect when the number of colors is not sufficient Dithering: randomly perturb the color values slightly to break up the contour effect
fixed pattern dithering diffusion dithering (the perturbed value of the next pixel depends on the previous one) Developed originally for rendering gray scale image using black and white ink only
Original value (R,G, or B) Dithered value

These colors are those that can be rendered consistently by different computer systems. They are obtained by quantizing the R,G,B component independently using uniform quanitization. Each component is quantized to 6 possible values: 0(0x00), 51(0x33), 102(0x66), 153(0x99), 204(0xCC), 255(0xFF).
Fall 2004 EL512 Image Processing Lecture 8, Page 31 Fall 2004

Dithering value
EL512 Image Processing Lecture 8, Page 32

Example of Color Dithering

Demo Using Photoshop

Show quantization results with different methods using image->mode->index color

Palette System (Windows) 256 colors

8 bit uniform without dithering

Fall 2004 EL512 Image Processing

8 bit uniform with diffusion dithering

Lecture 8, Page 33 Fall 2004 EL512 Image Processing Lecture 8, Page 34

Pseudo Color Image

Why?
Human eye is more sensitive to changes in the color hue than in brightness.

Pseudo Color Display

Intensity slicing: Display different gray levels as different colors
Can be useful to visualize medical / scientific / vegetation imagery
E.g. if one is interested in features with a certain intensity range or several intensity ranges

How?
Use different colors (different in hue) to represent different image features in a monochrome image.

Frequency slicing: Decomposing an image into different frequency components and represent them using different colors.

Fall 2004

EL512 Image Processing

Lecture 8, Page 35

Fall 2004

EL512 Image Processing

Lecture 8, Page 36

Intensity Slicing
Color C4

Example

C3 C2 C1

f0 =0

f1

f2

f3

f4

Gray level

Pixels with gray-scale (intensity) value in the range of (f i-1 , fi) are rendered with color Ci
Fall 2004 EL512 Image Processing Lecture 8, Page 37 Fall 2004 EL512 Image Processing Lecture 8, Page 38

Another Example

Pseudo Color Display of Multiple Images

Display multi-sensor images as a single color image
Multi-sensor images: e.g. multi-spectral images by satellite

Fall 2004

EL512 Image Processing

Lecture 8, Page 39

Fall 2004

EL512 Image Processing

Lecture 8, Page 40

An Example

Example

Fall 2004

EL512 Image Processing

Lecture 8, Page 41

Fall 2004

EL512 Image Processing

Lecture 8, Page 42

Color Image Enhancement

Enhance each primary color component independently using the techniques for monochrome images
Will change the color hue of the original image

Example of Color Image Enhancement

Convert the tri-stimulus representation into a luminance / chrominance representation, and enhance the contrast of the luminance component only.
Use HSI representation, where I truly reflects the luminance information.
Fall 2004 EL512 Image Processing Lecture 8, Page 43 Fall 2004 EL512 Image Processing Lecture 8, Page 44

Example of Color Image Enhancement

Example of Color Image Enhancement

Fall 2004

EL512 Image Processing

Lecture 8, Page 45

Fall 2004

EL512 Image Processing

Lecture 8, Page 46

Homework
1. (Computer Assignment) Write a program which first performs highpass filtering (you can use matlab func conv2 for this part) of an input gray scale image using the following filter: 0 1 0 Scale the filtered image to range between 0 and 255. Then displays the filtered image using 3 pseudo colors, using the following color transformation: Color Red for values 0-80, Color Green for values 81-160, color blue for values 161-255. In matlab, you can use the function colormap to change the colormap and use imshow to display an image using a specified colormap. Comment on the visual effect, e.g. each color represents what attributes of the image? (Computer Assignment) Choose a 24 bit RGB color image, perform the following operations: 1) convert it to YIQ format, save the resulting images in three separate files (for Y, I and Q components respectively), each with 8 bits/pixel; 2) perform histogram equalization to the Y image; 3) convert the enhanced Y image and the original I and Q image back to the RGB image. View the original and enhanced color RGB images and comment on your observations. (Computer Assignment) Choose a 24 bit color RGB image, quantize the R, G, and B components to 3, 3, and 2 bits, respectively, using a uniform quantizer in the range 0-256. Display the original and quantized color image using the original colormap associated with the image. Comment on the difference in color accuracy. Make sure you use a computer that has a 24 bit color display, and the test image has good color contrast.
EL512 Image Processing Lecture 8, Page 47
1 1 4 1 4 0 1 0

Reading
Prof. Yao Wangs Lecture Notes, Chapter 6. R. Gonzalez, Digital Image Processing, Chapter 6. A. K. Jain, Fundamentals of Digital Image Processing, Section 3.7 ~ 3.11, 7.7 ~ 7.8.

2.

3.

Fall 2004

Fall 2004

EL512 Image Processing

Lecture 8, Page 48