1. COMS 161 Introduction to Computing
Title: Digitizing Images
Date: February 21, 2005
Lecture Number: 16

2. Announcements
Homework 4
Due 2/25/05

3. Review
Digitization
Sound
Sampling
Quantizing

4. Outline The nature of images How digital images are
Natural vs. artificial images
How digital images are
Organized
Created
Stored
Processed

5. The Nature of Images Natural images Artificial images
From common, analog sources
Photos, drawings, paintings, TV, movies, etc.
Must be digitized for use with a computer
Artificial images
Generated digitally

6. Representing Digital Images
Natural images (such as a photograph, a frame of a video, etc.) typically consist of continuous or analog signals
Digital images are composed of pixels (picture elements)
For use in a computer, natural images must be digitized

7. Paint by Numbers Digital images are similar to paint-by-number kits 12 3

8. Paint by Numbers
Digital images are divided into a grid of square regions called pixels
Real images do not have coloring book areas of uniform color

9. Paint by Numbers Even within a pixel the region will not be uniform 12 3 4 5

10. Example of Digitization
Consider a photograph of a penny
Pretend that this is a photograph
To use this image in a computer, it must first be digitized

11. Example of Digitization
The first step in digitizing this natural image is sampling
This image is partitioned (sampled) into a 50×50 square grid of pixels
The picture resolution of this digitized image will thus be 50×50

12. Example of Digitization
An image’s aspect ratio is the ratio of the number of horizontal pixels to the number of vertical pixels
This 50×50 grid has an aspect ratio of 1:1
Most computer screens are 1.33:1
(640×480, 1024×768, etc.)
Std. TV is 4:3 (or 1.33:1)
HDTV is 16:9 (or 1.78:1)

13. Example of Digitization
The second step in digitizing the image is quantizing the pixels
For each pixel, an average color is calculated
This resolution (50×50) is ‘clearly’ insufficient to represent the detail of the original image

14. Resolution
Picture resolution is a trade-off between image quality and file size
This digitized image has a resolution of 272×416
Minimum file size is then (272×416) × (bytes/pixel)
For 256 colors (one byte per pixel), minimum file size would be (272×416) × (1) = KB
For 16 million colors (3 bytes/pixel), it would be 331 KB

15. Resolution
With the resolution reduced to 136×208, the picture loses detail
File size is reduced to:
28.3 KB for 256 colors
84.8 KB for 16 million colors

16. Resolution
With the resolution further reduced to 68×104, the picture becomes almost unrecognizable
File size is greatly reduced to:
7.1 KB for 256 colors
21.2 KB for 16 million colors
With large pictures and high color requirements, file size becomes very important
Digital cameras can easily create single pictures larger than 1 MB

17. Quantizing Digital Images
Imagine a simple image: a bright object on a dark background
Sample the image as before
Consider just a single row of pixels across the center

18. Quantizing Digital Images
Assign number values to the pixels: 0 = ‘black’ 1 = ‘white’
Plot the values of the pixels on the center row
0.0
0.5
1.0
With this image, we only need two “colors”, black and white

19. Dynamic Range Most pictures are more complex than just black and white
To adequately represent an image, we need enough levels of quantization to achieve the desired picture quality
The range of values chosen for quantization is called the dynamic range of the digitized image

20. Dynamic Range Max value – min value Typically it is a power of 2
256 gray values = 28, 8 bits / pixel
How large should a dynamic range be?
Science says we can only distinguish between 40 different shades of gray!!

21. Dynamic Range Example
This is a grayscale image quantized to 256 levels of gray
0 = ‘black’
127 = ‘medium (50%) gray’
255 = ‘white’
Dynamic range is sufficient for use in this presentation
Clear detail in highlights and shadows

22. Dynamic Range Example The same image, now quantized to 16 levels
0 = ‘black’
7 = ‘medium (50%) gray’
15 = ‘white’
Dynamic range is acceptable
Detail somewhat reduced in highlights and shadows
False contours becoming apparent (especially on chin and cheeks)

23. Dynamic Range Example The same image, now quantized to 4 levels
0 = ‘black’
1 = ‘dark (67%) gray’
2 = ‘light (33%) gray’
3 = ‘white’
Dynamic range is marginal
Detail severely reduced
Shadows flattened
Extreme false contouring

24. Dynamic Range Example The same image, now quantized to 2 levels
0 = ‘black’
1 = ‘white’
Dynamic range is unacceptable
Detail almost gone
But, this may be a desirable artistic effect

25. Dynamic Range Example