COMPUTER VISION 600.461

1
COMPUTER VISION600.461

• Larry Wolff
• MTW 11-12
• wolff_at_cs.jhu.edu, Office 212NEB
• Office Hours Wed. 1-2PM

2
REQUIREMENTS

• This is an introductory course on Computer
  Vision.
• No previous knowledge of computer vision or image
  analysis is required.
• Prerequisite is only that you be computer
  literate
• -Know how to program in C.
• -Have had some exposure to UNIX type
• environments.
• -Able to use a Web Browser.

3
GRADING

• Midterm 20
• Final 30
• Problem Sets and Lab Projects 50

4
RESOURCES

• You will need an account on a workstation such as
  a SUN, DEC, or an SGI. Of course direct access to
  the JHU network is highly desireable.
• PCs running Linux may be OK. Windows could be
  problematic.

5
OVERVIEW
COMPUTER IMAGING
COMPUTER VISION
IMAGE PROCESSING
6
(No Transcript)
7
X
X
X
VOLTAGE
X
TIME
One Scanline
One Pixel
IMAGE
8
ORGANIZATION OF A 2D IMAGE
Pixel Binary 1 bit Grey 1 byte Color 3 bytes
9
BINARY IMAGE
10
GREYSCALE IMAGE
11
COLOR IMAGE
12
SCHEMES FOR REPRESENTING COLOR
RED
INTENSITY
red
magenta
magenta
yellow
saturation
yellow
hue
blue
green
cyan
BLUE
GREEN
cyan
HSL
RGB
13
IMAGE FILE FORMATSWhy are there so many ?
BMP
PPM
tiff
JPEG
pgm
gif
EPS
14
(No Transcript)
15
OTHER IMAGING MODALITIES

• Medical Imaging
• Range sensing
• Thermal IR

16
CATSCAN IMAGE
17
RANGE IMAGE
18
THERMAL IMAGE
19
THE HUMAN EYE
20
THE HUMAN EYE
21
HIERARCHICAL IMAGE PYRAMID
High Level
Feature Extraction
Segmentation Edge Detection
(spectrum, edges, lines)
Preprocessing (images, subimages)
Raw Image Data (pixels)
Low Level
22
CONTRAST ENHANCEMENT
23
THRESHOLDED SEGMENTATION
24
HOMEWORK ASSIGNMENTDue Wed. Sept. 16

• Get familiar with a really useful image display
  tool called XV
• Works on popular Command Line environments such
  as UNIX and on DECs and SGIs.
• If you dont already have this you can ftp from
  ftp//ftp.cis.upenn.edu/pub/xv

25
HOMEWORK ASSIGNMENTDue Wed. Sept. 16

• Access image files by logging into cs.jhu.edu via
  anonymous ftp
• Display the image trees.gif and enhance contrast
  using intensity correction.
• Try to segment trees.gif into a binary image
  using intensity correction.
• Display desert.gif and modify individual RGB
  color bands.