Computer Science 111

1
Computer Science 111

• Fundamentals of Programming I
• Introduction to Digital Image Processing

2
Digital Images

• Input devices
• scanners
• cameras
• camcorders
• Output devices
• display screens
• printers
• Processing
• file compression
• various transformations

3
Transformations
Convert from color to grayscale Adjust the
brightness Adjust the contrast Adjust the
size Rotate Morph into another image
4
Morph of the Day
5
Representing Images

• An image can be represented as a two-dimensional
  grid of RGB values (pixels)
• To capture an image (via camera or scanner), a
  continuous range of color info is sampled as a
  set of discrete color values
• All processing works with the grid of RGB values
• Output maps the grid to a display or a printer

6
The images Module

• A non-standard, open source module that includes
  a set of classes and methods for processing
  images
• Can edit scripts in IDLE, then run them from IDLE
  or a terminal prompt

python testimages.py
7
The Image Class
Image(fileName) Image(width, height) draw() clo
ne() getWidth() getHeight() getPixel(x,
y) setPixel(x, y, (r, g, b)) save(ltoptional
file namegt)
Represents a grid of pixels Methods for display,
examining the dimensions, examining or resetting
pixels, and saving changes to a file A pixel is
just a tuple of 3 integers
8
A Simple Session
gtgtgt from images import Image
Uses a terminal prompt to launch python Import
the relevant class from the module The images
library must be in the current working directory
9
A Simple Session
gtgtgt from images import Image gtgtgt image
Image('smokey.gif')
The image file must be in the current working
directory
10
A Simple Session
gtgtgt from images import Image gtgtgt image
Image('smokey.gif') gtgtgt image.draw()
draw must be run to display the image
11
A Simple Session
gtgtgt from images import Image gtgtgt image
Image('smokey.gif') gtgtgt image.draw() gtgtgt
image.getWidth(), image.getHeight() (300, 225)
The image window must be closed to continue
testing The comma creates a tuple of results
12
A Simple Session
gt python gtgtgt from images import Image gtgtgt image
Image('smokey.gif') gtgtgt image.draw() gtgtgt
image.getWidth(), image.getHeight() (300,
225) gtgtgt image.getPixel(0, 0) (206, 224, 122)
13
Transformations Black and White

• Compute the average of the three color components
  in a pixel
• If the average is less than 128, then set the
  pixels three color components to 0s (black)
• Otherwise, set them to 255s (white)

14
Transformations Black and White
blackPixel (0, 0, 0) whitePixel (255, 255,
255) for y in xrange(image.getHeight())
for x in xrange(image.getWidth()) (r,
g, b) image.getPixel(x, y) average
(r g b) / 3 if average lt 128
image.setPixel(x, y, blackPixel)
else image.setPixel(x, y,
whitePixel)
15
Transformations Inversion

• Should turn black into white or white into black
• Reset each color component of a pixel to 255
  minus that components value

16
Transformations Inversion
for each pixel in the image subtract each
color component from 255
17
Transformations Grayscale

• Compute the average of the three color components
  in a pixel
• Reset each color component of the pixel to this
  average value

18
Transformations Grayscale
for y in xrange(image.getHeight ()) for x in
xrange(image.getWidth()) (r, g, b)
image.getPixel(x, y) ave (r g b) /
3 image.setPixel(x, y, (ave, ave, ave))
19
A Better Grayscale Algorithm

• The simple average of the RGB values does not
  take account of the human retinas different
  sensitivities to the luminance of those values
• The human eye is more sensitive to green, then
  red, and finally blue
• Psychologists have determined the exact
  sensitivities
• Multiply each value by a weight factor and then
  add them up

20
A Better Grayscale Algorithm
Old

• red int(red 0.299)
• green int(green 0.587)
• blue int(blue 0.114)
• gray red green blue
• image.setPixel(x, y, (gray, gray, gray))

New
21
Package Code in a Function

• def grayScale(image)
• for y in xrange(image.getHeight ())
• for x in xrange(image.getWidth())
• (r, g, b) image.getPixel(x, y)
• ave (r g b) / 3
• image.setPixel(x, y,
• (ave, ave, ave))

Note that this function does not return a new
image, but modifies its argument Until now,
functions did not modify arguments This is the
most efficient way of modifying large data
objects
22
For Wednesday

• Finish the reading the handout on image processing