1. Digital Image Fundamentals
Chapter 2

2. KEY CONCEPTS Sampling and quantizing in digitizing images
Pixels and image resolution
Image bit depth

3. KEY CONCEPTS
How pixels, image resolution, and bit depth are related to sampling and quantizing
Color representation in digital images
Bit-mapped images versus vector graphics

4. Natural Image (pretend)

5. Capture: Scanner or Digital Camera

6. Sampling How many pixels do you capture?
Area the might have much detail
Is reduced to one pixel
“Averaging” effect

7. Sampling Each Sample Point is translated into a pixel
Real world image: Infinite detail
Single Pixel: Finite detail

8. Quality Loss illuminated
Once the digital image is captured….
There is no way to “restore” the original detail.

9. Analog / Real World Images
In the real world, we can perceive almost infinite detail
By…
Moving closer
Using a microscope
Nano-visualization

10. Sampling: Number of Pixels
Original
7500 pixels (100 X 75)

11. Sampling: Number of Pixels
Original
1900 pixels (50 X 38)

12. Sampling: Number of Pixels
Original
300 pixels (20 X 15)

13. Sampling Scanners Digital Cameras Up to 4800 pixels per inch
8 X 10” photograph 
38,400 pixels X 48,000 
1,843,200,000 
1.7 Gigapixels * 32-bit color
6.8 Gigabytes
Up to 12 Megapixels
4000 X 3000 pixels
Consider 300dpi printing
13 X 10’ looks great
Consider 150dpi printing
26 X 20” looks pixelated.

14. Megapixel Associated with digital cameras
Camera manufactures do NOT follow the base-2 standard so
a Megapixel is literally 1 million pixels
not 220 pixels.
A 3000 × 2000 pixel digital image is 6,000,000 pixels
Referred to as six megapixels.

15. Pixels and Dots Pixel Dot Just the base unit of storage.
Does NOT have a set size, i.e., 1 millimeter X 1 millimeter
Pixel size depends on the monitor or projector
May not be a square
Could be a circle of light
Smallest unit that a printer can print.
Smallest unit that a scanner can detect

16. Over sampling Scanner Printer 4800dpi 4800/300 = 16 16 X 16 300dp
Printed as one dot

17. Under sampling Scanner Printer 300dpi One dot 600dpi 600/300 = 2
Printed using 2 X 2
NOT taking full advantage of the printers capabilities.

18. Quantization: Number of Colors
Original (16.7 million colors)
2 colors (1-bit)

19. Quantization: Number of Colors
Original (16.7 million colors)
4 colors (2-bit)

20. Quantization: Number of Colors
Original (16.7 million colors)
8 colors (3-bit)

21. Quantization: Number of Colors
Original (16.7 million colors)
16 colors (4-bit)

22. Quantization: Number of Colors
Original (16.7 million colors)
32 colors (5-bit)

23. Quantization: Number of Colors
Original (16.7 million colors)
64 colors (6-bit)

24. Quantization: Number of Colors
Original (16.7 million colors)
256 colors (8-bit)

25. Quantization A natural image is colored in continuous tones
theoretically has an infinite number of colors.
Binary representation restricts the reproduction colors and shades.
8 bits can only encode 256 different color values
In image capturing, the process of encoding an infinite number of possible colors into a finite number list of colors is called quantization.

26. Color selection
Quantization limits how many colors you can represent in a digital image, but
Not which colors you decide to encode.
Although this image has only 8 colors
The colors are very specific to the original image.

27. Color Palettes
Image formats (like GIF) can encode different palettes - depends on colors in the original image.
If you chose the colors wisely the image looks more realistic.

28. Color Translation
Meta data “maps” each 3-bit code to a universal 32-bit palette 000        

29. Red Green Blue
A single 8-bit byte can be used to represent one of 256 different possible values.
The values range from 0 to 255.
An RGB (red, green, blue) color can be represented in three 8-bit bytes
Example:
Red: 255 (maximum red) 8-bits
Green: 0 (minimum green) 8-bits
Blue: 127 (50% blue) 8-bits 24-bits total
What color is this?   

30. Red Green Blue
Although 24-bit color is sufficient for human vision, 48-bit RGB can be used when applications need to analyze images beyond the spectrum visible to humans.
48-bit RGB color is represented using 16 bits per component of R, G, and B.

31. 32-bit color
A 32-bit image is basically 24-bit RGB with an additional 8-bit alpha channel.
The alpha channel is used to specify the level of transparency.
Unlike 24-bit images that are fully opaque, 32-bit images can be smoothly blended with other images.

32. Alpha-channel Refers to the transparency of a color.
Transparency is a powerful feature because it better models real world entities.
Consider a digital image of a tinted window.

33. Questions
How many possible colors can be represented with 48-bit color depth?
How many possible levels of red can be represented with 48-bit RGB color?
How many times would a file size increase by going from 24-bit to 48-bit?

34. Bitmapped Images
Hopefully, lab helped illustrate this…

35. Bitmapped Images
But now consider that instead of these being 1’s and 0’s, they are 24-bit, 32-bit, or 48-bit color codes.

36. Bitmaps  Raster Graphics
Bitmapped images also are called raster graphics
“rastering” refers to the way most video displays translate the images into a series of horizontal lines on the screen.  

37. Bitmaps
Bitmap image formats are the most commonly used in image-editing applications.
However, bitmap appearance depends on the resolution of the output device
Bitmapped images can appear jagged and lose detail when they’re scaled onscreen or printed.

38. Jagged images (or Jaggies)
A picture is worth a thousand words, right?

39. Alternative to Bitmaps
Besides pixel by pixel representation, what other way could we store images digitally?
Hmmm? Think about it.
Math Power!

40. Vector Graphics
Vector graphics is the use of geometrical primitives such as
points,
lines,
curves, and
polygon(s)
…to represent images.

41. Vectors x2, y2 points, lines, curves, and polygon(s)
are all based upon mathematical equations.
x, y coordinate
x1, y1

42. Splines (not in book)
A spline is a curve defined by piecewise polynomial functions.

43. Fonts – The first vector images
vectors still appear smooth at higher magnification
rasterized graphic
rasterized graphic with anti-aliasing

44. Fonts
Font magnification example: g

45. Rasterization Displayed as scalable raster Stored as vector
Vectors are literally made up of mathematical formulas
No pixels at all
In principle, vectors can be rendered at limitless resolution
Monitors and projects still display using pixels.
To display a vector image, software has to convert the vector information into a temporary raster image.
Called rendering or rasterization

46. Aliasing (formal word for jagged)
Rasterized images will always appear jagged
You just have to zoom in.
This jagged effect is called aliasing
caused by under-sampling or over-maginfication.

47. Anti-Aliasing
Pixels with intermediary shades can be used to soften the jaggedness
This technique is called anti-aliasing.
Technique to make rasterization more smooth.

48. Raster (bitmap) and Vector formats
GIF
JPEG
PICT
TIFF
SWF (Shockwave Flash)
SVG (Scalable Vector Graphic)
EPS (Encapsulated Postscript)
AI (Adobe Illustrator)
Combo-format (not in book)
PNG (Portable Network Graphic) – Raster format but includes vector information when applicable.

49. File Size
How much file space does a 6-megapixel 24-bit color image take up uncompressed.
6,000,000 pixels × 24 bits per pixel
= 144,000,000 bits
File size in bytes:
144,000,000 bits/(8 bits per byte)
= 18,000,000 bytes

50. Obvious Compression Reduce sampling Lower the color depth
Lower the resolution
3000 X 2000 (6 megapixels)  1500 X 1000 (3 megapixels
Lower the color depth
24-bit color  16-bit color
Maybe there are only 20,000 different colors present

51. Clever Compression Run-length encoding (RLE)
runs of data are stored as a single data value and count, rather than as the original run.
(25 bits)
(0,7) (1,5) (0, 13)
(0, 0111) (1, 0101) (0, 1101)
(15 bits)
Patterns instead of runs can be counted.

52. Clever Compression Huffman coding
Find most frequent color, use smallest representation
1000 white pixels (1)
80 black pixels (01)
24 pink pixels (001)
8 red pixels (0001)
3 blue pixels (00001)

53. LZW Compression
Lempel-Ziv-Welch (LZW) is a universal lossless data compression algorithm
created by Abraham Lempel, Jacob Ziv, and Terry Welch.
It was published by Welch in 1984
Used in GIF compression.

54. DEFLATE Compression
Deflate is a lossless data compression algorithm that uses a combination of the LZW algorithm and Huffman coding.
Used by the PNG format.

55. JPEG The name "JPEG" stands for Joint Photographic Experts Group,
the name of the committee that created the standard.
Standardized format published in 1992
most widely-used format for storing digital photographs
Lossy compression algorithm
Reduces pixel blocks (typically 8X8) into a single representation.

56. Representing Color

57. Color is a light wave
The wavelengths of visible light range from about 380 to 700 nm (nanometers)—creating a continuous spectrum of rainbow color, from the violet end (380 nm) to the red end (700 nm).

58. Color is a light wave
The retina of the human eye has two categories of light receptors: rods and cones.
Rods are active in dim light but have no color sensitivity.
Cones are active in bright light and have color sensitivity.

59. Color is a light wave
There are three types of cones: one type is sensitive to red (ρ) wavelengths, one to green (γ), and one to blue (β). The curves represent the relative sensitivity of these three receptors for the normal human eye

60. What is color?
Color is how our eyes perceive different forms of energy.
Energy moves in the form of waves.

61. What is a wave?
Think of a fat guy (Dr. Breimer) doing a cannonball into a pool.
The incredible energy created by my fat ass hitting the water is transfer and dispersed into the pool in the form of a wave

62. Why does energy move in waves?
I don’t f***ing know. Are you 4-years old? you have to ask a million stupid questions?
Seriously, there is some complex physics behind the reason, but here is a simple way to explain it….

63. Why does energy move in waves?
Q: How does a snake move without legs?
A: By going “swish swish”
Similarly, the “swish swish” of a wave allows energy to move even in a vacuum.

64. Why does energy need to move anyway?
To get a 40oz beverage from the liquor store

65. Where were we anyhow?
Light is a form of energy that travels in a wave pattern.
The length of the wave can vary
Short wavelength
Long wavelength

66. The Human Eye…
has Cones and Rods (like nerves) that can detect different wavelengths of light…
and send signals to the brain.

67. Visible Energy We can only see a very limited range of wave lengths.
What would it be like if we could see microwaves?

68. What microwaves might look like

69. Spectrum of visible light

70. What is your favorite color?
Can you guess mine?
Infared
My son’s favorite color is yellow, red, black, white, blue, purple, brown (poop color), khaki (light poop color), and orange.
This is his way of saying he hates pink

71. Tristimulus Theory
Any color can be produced by mixing different amounts of three additive primaries

72. Monitors
Create color by using a combination of red green and blue light (RGB)

73. How do TVs and Computer Monitors create color?

74. How do TVs and Computer Monitors create color?
The same way our eyes detect color.
By mixing the three wavelengths your eyes can detect.

75. Red, Yellow and Blue (NOT!)
In kindergarten, we all learned that the primary colors were:
Red, Yellow, and Blue, right?
Well, that was a lie.
Just, like in 1st grade when they told you there was a giant vacuum in space.
There is NO giant vacuum in space.
Microwaves are NOT invisible.
And, Yellow is NOT a primary color!

76. Yellow Yellow is ONLY considered primary when mixing paint or ink
Mixing paint is different than mixing light
More colors = darker color
Red + Green is too dark (brownish, not yellow)

77. Green Mixing light is different than mixing paint.
It is an additive and synergistic process
More color = lighter color
Red + Green = bright yellow.
Red + Green + Blue = white!

78. Back to TVs and Monitors
The surface is black, no light equals black.
Each pixel is created from three separate light signals.
Two models:
RGB: Red, Green, Blue
CMYK:
Cyan
Magenta
Yellow
Key (level of intensity – bright to dark)

79. Pixel Components
If you put colors close enough together, the eye perceives them as one color.

80. TVs and Monitors Light signals can be generated in many different ways
The key is that you want the pixel to be very small and bright.
Three technologies:
CRT: Cathode Ray Tube
LCD: Liquid Crystal Display
Plasma

81. CRT: Cathode Ray Tube
Glass tube containing an electron gun and a fluorescent screen

82. LCD: Liquid Crystal Display
Each pixel consists of a layer of molecules aligned between transparent electrodes, and polarizing filters

83. Plasma TV
Cells between two panels of glass hold neon and xenon gas. Gas is electrically turned into a plasma which excites phosphors to emit light.

84. RGB vs. Wavelength
Technologically, it is easier to control color by emitting three different colors RGB, rather than vary the wavelength to create a “pure” color.
Similar to Binary
Can encode any number in binary
Can encode any color with RGB combination

85. RGB vs. Wavelength
In fact, the cones and rods in the eyes detect only three colors.
We see more than three because the cones and rods send “mixed” or synergistic signals to the brain.
Humans have a hard time distinguishing RGB mixtures from “pure colors” because we sense color as RGB mixtures anyway.

86. RGB is great but not perfect
You can NOT reproduce all the visible color wavelengths using RGB combinations
But, you can get pretty close.

87. RGB vs. CMYK RGB is NOT suitable for printing on paper.
Color printers can NOT produce Yellow (Red+Green) because ink does not have the same synergistic properties of light.
Thus, Yellow has to be a primary pigment.
The color wheel gets turned.

88. RGB vs. CMYK
CMYK: Cyan, Magenta, Yellow, and K (Key) which is really black.
RGB is used almost exclusively for TVs/Monitors (where the surface is Black), you don’t need Key/Black
Because CMYK is also for print (where paper is typically white), you need Black (C+M+Y = purplish brown).
How do you get White with RGB?

89. Complementary Colours
170–171
Complementary Colours
Subtract additive primary from white gives its complement
Equivalently, add other two additive primaries
C = G+B = W-R
M = R+B = W-G
Y = R+G = W-B
Cyan, magenta and yellow are subtractive primary colours (mixing ink/paint)

90. CMYK
CMYK encoding is used for applications that focus on printing: Photo Developing software and publishing software like QuarkXpress, Framemaker, etc.
Applications that use RGB must convert to CMYK for printing
Some RGB colors (on the monitor) can be perfectly matched using CMYK.

91. RGB vs. CMYK

92. Digital Color
Operating Systems and applications encode color using bits.
Very early color systems only used 2 bits (4 colors).
Dr. B’s first computer (IBM 8086) supported only 4 colors CMYK.
As process speeds increased and graphics hardware improved
8 bit color and 16 bit color became the standard ( )

93. Data  Color Assume a four color encoding (2 bits)
Assume a monitor with 640 X 480 pixels
Monitor refreshes 60 times per second
(60 Hertz)
The operating system must send…
640 X 480 X 2 X 60 bits per second.
= 36 million bit per second.

94. Data  Color: Hardware Monitor plugs into a video/graphics card.
The video card converts the bit pattern into an electrical signal.
Monitors and graphics cards work together because of international standards.
For example, VGA standard

95. Monitor The electrical signal triggers the pixel color.
CRT and LCD technology has a limit on
How small a pixel can be.
How bright it can be
How often it can be refreshed
60-90 Hertz is the typical range

96. Data  Color: Software
The graphics card actually plugs into the mother board of the computer.
The bit pattern travels across the motherboard.
A device driver is used so that the operating system can communicate with the graphics card.
A device driver is just small program…still written directly in assembly language.

97. Graphics Cards Old graphics card were just signal converters
New graphics cards have memory (RAM) and processors
Takes the burden off of the computer’s processor.
Enables 24-bit color at resolutions as high as 2560x1600.
Plus graphics card can also do things like render vectors (geometry computations).

98. RGB Color Depth Choose number of bits for each of R, G and B
More bits per color means more total colors, but image files will be larger
8 bits per color is not the standard: 24-bit color, 16.7 million colors
255
218

99. RGB Color Depth
8 bits (1 byte) per component means that you have 256 different “levels”
If R = G = B, color is a shade of gray.
Human eye can distinguish 256 shades of gray
So, while 16.7 million colors is beyond what the human eye can distinguish.
24-bit RGB is under quantized for gray.
But for Gray only.

100. Practical Technique: Color Palettes
Choose 256 most important colors in an image to store in its palette
When 24-bit image is reduced to indexed color, some colors may be missing form the palette
Replace missing color by nearest, may lead to posterization
Dither – use pattern of dots and optical mixing
Web-safe palette – 216 colors guaranteed to reproduce accurately on all platforms and browsers

101. HSV Alternative way of specifing colour
173–176
HSV
Alternative way of specifing colour
Hue (roughly, dominant wavelength)
Saturation (purity)
Value (brightness)
Model HSV as a cylinder: H angle, S distance from axis, V distance along axis
Basis of popular style of colour picker