PCjs Machines
Home of the original IBM PC emulator for browsers.
PC-SIG Diskette Library (Disk #1678)
[PCjs Machine "ibm5170"]
Waiting for machine "ibm5170" to load....
Information about “EYESIGHT”
Enter the world of artificial intelligence and learn about computer
vision with EYESIGHT.
EYESIGHT is a vision processing program designed as an introduction to
the field of computer vision. You do not need a camera or any prior
knowledge of computer vision to use EYESIGHT. EYESIGHT provides an
environment for reading and writing image files, performing various
image processing functions, and displaying the results.
Computer vision is a branch of artificial intelligence attempting to
give computers the gift of sight. Using EYESIGHT, you will learn how
this is possible. EYESIGHT comes with sample image files that you can
use to experiment and learn with. You will learn about region
recognition and edge detection. EYESIGHT provides seventeen different
image processing functions with which you can experiment. Also included
is a tutorial on computer vision. EYESIGHT is an exciting look into
computer vision.
FILE1678.TXT
Disk No: 1678
Disk Title: EyeSight
PC-SIG Version: S1
Program Title: EyesSight
Author Version: 1.7
Author Registration: $25.00
Special Requirements: CGA.
Enter the world of Artificial Intelligence and learn about Computer
Vision with EYESIGHT.
EYESIGHT is a vision processing program designed as an introduction to
the field of computer vision. You do not need a camera or any prior
knowledge of computer vision to use EYESIGHT. EYESIGHT provides an
environment for reading and writing image files, performing various
image processing functions, and displaying the results.
Computer vision is a branch of Artificail Intelligence attempting to
give computers the gift of sight. Using EYESIGHT, you will learn how
this is possible. EYESIGHT comes with sample image files that you can
use to experiment and learn with. You will learn about pixel
recognition and edge detection. EYESIGHT provides seven different image
processing functions that you can experiment with. Also included is a
tutorial on computer vision. EYESIGHT is an exciting look into computer
vision.
PC-SIG
1030D East Duane Avenue
Sunnyvale Ca. 94086
(408) 730-9291
(c) Copyright 1989 PC-SIG, Inc.
GO.TXT
╔═════════════════════════════════════════════════════════════════════════╗
║ <<<< Disk #1678 EYESIGHT >>>> ║
╠═════════════════════════════════════════════════════════════════════════╣
║ To start program, type: EYESIGHT (press enter) ║
║ ║
║ To print documentation, type: COPY MANUAL.DOC PRN ║
╚═════════════════════════════════════════════════════════════════════════╝
MANUAL.DOC
E Y E S I G H T
V E R S I O N 1.7
A ShareWare Computer Vision Program for IBM PCs
and compatibles
Copyright (c) 1989 by Robots, Etc
Loren Heiny
Robots, Etc
P.O. Box 122
Tempe, AZ 85280
T A B L E O F C O N T E N T S
WHAT IS EyeSight?.......................................1
WHAT CAN EyeSight DO?...................................1
WHAT EQUIPMENT DO YOU NEED?.............................1
WHAT FILES ARE ON THE EyeSight DISK?....................1
WHAT IS COMPUTER VISION?................................1
REGISTERING YOUR COPY OF EyeSight.......................2
GETTING STARTED.........................................3
MAKE A WORKING COPY.....................................3
INSTALLING EyeSight ON A HARD DISK......................3
RUNNING WITH EyeSight...................................3
HOW DO I GET OUT OF HERE? or HELP!....................3
LET'S DO SOMETHING......................................4
TRYING TO THRESHOLD.....................................4
EXPERIMENT SOME MORE....................................5
USING THE ENVIRONMENT...................................5
CHANGING YOUR MIND......................................6
VISION PROCESSING SUPPORTED.............................6
CONFIGURABLE EXEC MENU..................................7
ALL IN AN IMAGE.........................................7
WHAT YOU SEE............................................7
THE LEFT DISPLAY WINDOW.................................7
THE RIGHT DISPLAY WINDOW................................8
IMAGE COLORS............................................8
SELECTING THE DISPLAY WINDOW............................8
COPYING BETWEEN DISPLAY WINDOWS.........................9
READING AN IMAGE FILE AND MULTIPLE IMAGE BUFFERS........9
READING A FILE FROM THE COMMAND-LINE...................10
LISTING THE IMAGE BUFFERS..............................10
WRITING AN IMAGE FILE..................................10
DEFAULT PATHNAMES......................................11
CONFIGURATION FILE.....................................11
HELP!..................................................12
INTERNAL DIRECTORY.....................................12
EXITING THE PROGRAM....................................12
PRINTING AN IMAGE......................................12
EXAMINING PIXELS.......................................13
TWO IMAGE OPERATIONS...................................13
IMAGE ANALYSIS FUNCTIONS...............................13
THE CONTOUR FOLLOWER...................................15
EXPERIMENTING WITH EYESIGHT............................16
SCRIPT LANGUAGE........................................19
SETTING THE GRAPHICS MODE..............................19
COMMAND SUMMARY........................................20
QUICK COMMANDS.........................................21
__________________________________________________________EyeSight
WHAT IS EyeSight?
EyeSight is a ShareWare vision processing program designed
as an introduction to the field of computer vision. You do
not need a camera or any prior knowledge of computer
vision to use EyeSight. (For a brief introduction to computer
vision refer to the file vision.doc.)
WHAT CAN EyeSight DO?
EyeSight provides an easy-to-use environment for reading and
writing image files, performing various image processing
functions, and displaying the results.
WHAT EQUIPMENT DO YOU NEED?
IBM PC/XT/AT or compatible computer
A graphics adapter: CGA, EGA, or VGA
At least 1 floppy drive (Hard disk recommended)
DOS Version 2.1 or later
At least 256K of memory (640k recommended)
WHAT FILES ARE ON THE EyeSight DISK?
manual.doc This file
vision.doc An introductory article on computer vision
eyesight.exe The image processing program
eyesight.hlp Internal help file used by EyeSight
clips.im An image of clips
computer.im An image of a computer
mtn.im An image of a mountain scene
bolts.im An image of bolts
balloon.im An image of some balloons
flower.im An image of a flower
demo.scr A script file which demonstrates
some of EyeSight's vision routines
cga.cfg Configuration file for CGA mode
ega.cfg Configuration file for EGA mode
vga.cfg Configuration file for VGA mode
WHAT IS COMPUTER VISION?
Computer Vision is a branch of Artificial Intelligence (AI)
attempting to endow computers with the gift of sight. There is
little doubt that for most of us vision is our most relied upon
sense. And as such, it is no wonder that computer vision is
receiving such a large amount of attention in the AI community.
- Page 1 -
______________________________________________________________EyeSight
Already computer vision systems can be found on industrial assembly
lines inspecting products, guiding robots, and reading identification
markers. Of course, there are numerous applications that go beyond
the doors of industry, such as text reading machines and vision
systems for household robots. And the applications are only beginning
to unfold.
However, up to this point, computer vision technologies have rarely
been designed for the vast numbers of personal computers available.
EyeSight is a break from this tradition. With EyeSight you will be able
to experiment with a few simple vision processing techniques and
along the way familiarize yourself with several of the basic
concepts involved.
REGISTERING YOUR COPY OF EyeSight
EyeSight V1.7 is a ShareWare program and in the tradition of
ShareWare, you are encouraged to try it out and share it with
your friends. However, if you use EyeSight regularly and
would like to encourage its further development, please
register your copy by sending $25 in check or money order to
the address below (AZ residents add sales tax). Please include
the version number of your copy of EyeSight.
If you register your copy of EyeSight you will enjoy the following
benefits:
1) One free version update of EyeSight
2) Technical support
3) A diskette of sample images from our image library
4) A one year subscription to the EyeSight newsletter -- an
occassional publication dedicated to the computer vision
experimenter which discusses ways to explore computer
and robot vision using EyeSight.
Please send all registrations and inquiries to:
Robots, Etc
P.O. Box 122
Tempe, AZ 85280
* IBM PC and IBM are registered trademarks of International
Business Machines Corporation
- Page 2 -
______________________________________________________________EyeSight
GETTING STARTED
The following sections will lead you through a step-by-step
introduction to EyeSight. Simply follow the instructions as they
are presented and you should have little trouble stepping into the
world of computer vision.
MAKE A WORKING COPY
Before you do anything else, it is a good idea to make a working
copy of the EyeSight diskette. Once you have done this, set aside
the original EyeSight diskette and use your working copy from
this point on.
INSTALLING EyeSight ON A HARD DISK
If you want to run EyeSight from your hard disk, simply create a
sub-directory and copy all of the files on the EyeSight diskette into
it. At this early stage in the game there isn't any reason to build
up an extensive directory structure for EyeSight. Why not just keep
it simple!
RUNNING WITH EyeSight
To run the program, get into the directory where the EyeSight.exe
file resides and simply type:
eyesight <Enter>
at the DOS prompt (Whenever you see <Enter>, press the key marked
Enter).
EyeSight will now load and its environment will appear on the
screen with an author message. Press any key to continue with the
program. (If you are using a VGA card and the majority of the
screen is purple, see the section SETTING THE GRAPHICS MODE.)
If your machine does not have enough memory the message:
Not enough memory
will appear. If you do not have a CGA,EGA, or VGA card, the program
will not run correctly either.
HOW DO I GET OUT OF HERE? or HELP!
You can exit EyeSight at any time by holding down the ALT key and
pressing the letter x. In addition, Help is accessible from most
points in the program by pressing the F1 function key.
- Page 3 -
______________________________________________________________EyeSight
LET'S DO SOMETHING
The EyeSight distribution disk includes several images that are ready
for you to use. In this section we'll read one of these image files
into memory and then perform an image processing operation on it.
First let's read in one of the image files supplied on the EyeSight
disk. The read function is a file operation and is listed with the
other file options under the "File" menu option. To load an image:
1) Use the left and right arrow keys on the number
pad to move the highlighted section in the main
menu bar to the "File" option.
2) A pull-down menu will appear with several options.
Select "Load Image" by pressing the Enter key.
3) A prompt will appear for the filename of the
image to read. Type:
clips.im <Enter>
This is an image of several dark plastic clips on
a white background.
4) The image of the clips should now appear on the
left side of the screen.
Now let's try applying one of the vision processing functions to the
image. You'll find several image processing routines located in
the "Exec" pull-down menu. To get to these options, press the left
or right arrow keys until "Exec" is highlighted in the main menu bar.
A pull-down menu will appear with the options:
Histogram
Profile
Threshold
Roberts
Average
Contour
Burns
TRYING TO THRESHOLD
You can select any of these image processing functions, but first
let's try thresholding the image first. Never mind if you aren't
sure what this really does - for now just try to get used to the
environment (The section called IMAGE PROCESSING FUNCTIONS will
discuss what each of these functions do). To select the thresholding
function:
1) Move the up-down arrow keys until the "Threshold"
option is highlighted. Press the <Enter> key.
- Page 4 -
______________________________________________________________EyeSight
2) A slide-bar will appear in the dialogue box at the
bottom of the screen. The slide-bar is used to
select a threshold value that will be used in the
thresholding operation. For now just use the default
value by pressing <Enter>.
A thresholded image, which is simply a black-and-white version of
the original image, will now appear on the right side of the
screen.
Congratulations, you have just completed your first image
processing operation with EyeSight! It's easy, isn't it?
EXPERIMENT SOME MORE
While you have this image in memory, why not try thresholding the
image again, but this time try using another threshold value. To
perform another thresholding operation, select the "Threshold"
function as before, but this time when the threshold bar pops-up
in the dialogue box, use the left and right arrow keys to reposition
the slide-bar to another threshold value. When the slide-bar is
where you want it, press <Enter> and the thresholding operation
will use this new threshold value.
If you would like to go through a more in-depth example of using
EyeSight, look at the section "Experimenting with EyeSight" which
appears at the end of this manual.
USING THE ENVIRONMENT
EyeSight provides an extensive environment for vision processing. The
functions are built around pull-down menus that you can easily
access with the cursor keys, or as you become more familiar with
the program, through specially defined keys.
The various ways of pulling down a menu are:
1) Move the highlighted bar to the desired option
with the arrow keys.
or 2) If no pull-down menus are present, you can press
the first upper case letter of the desired
command. For example, you can type 'F' to select
the "File" option.
or 3) While in any pull-down menu you can jump to
another pull-down menu by holding down the ALT
key and pressing the first upper case letter of
the desired command.
or 4) If the highlighted bar is on the selection and
no pull-down menu is present, press the "Enter"
key.
- Page 5 -
______________________________________________________________EyeSight
Ways of selecting a function within a pull-down menu:
1) Move the highlighted bar up and down using the
arrow keys until the desired function is high-
lighted, then press the <Enter> key.
or 2) Type the capitalized letter that is in the
desired function name (i.e. e, f, s, p, o, or x).
or 3) Some common functions, like reading an image file
have special function keys that can be used from
many points in the program (for a listing of
these see the section "Quick Commands").
CHANGING YOUR MIND
To abort or exit ANY command or pull-down menu use the "ESC" key.
VISION PROCESSING SUPPORTED
The following image processing functions are provided by default
in EyeSight.
Function Brief Description
========================================================
Histogram Display a histogram of the current
image buffer
Profile Display the profile of the image
line specified for the current
image buffer
Threshold Threshold the current image buffer
Roberts Calculate and display the roberts
gradient of the current image
buffer
Average Average the current image and
display the result
Contour Trace the contours of regions in a
black and white image
Burns Locates lines in the current image
In addition, there are several other functions that you can add
to the EXEC menu. These functions are included under the OPTIONS/
ADD FUNCTION selection. A list of these routines are:
Function Brief Description
========================================================
Sobel The Sobel edge detector
Prewitt The Prewitt edge detector
Sharpen Enhance the edges of the image
Contrast Stretch the contrast of the image
And Bitwise and two images
Or Bitwise or two images
- Page 6 -
______________________________________________________________EyeSight
Function Brief Description (Continued)
========================================================
Add Add two images pixel by pixel
Subtract Subtract two images
Invert Invert image values
Log Takes the log of the image
CONFIGURABLE EXEC MENU
These functions can be accessed by first adding them to the EXEC
pull-down menu. This is accomplished by selecting the OPTIONS/
ADD FUNCTION menu choice. A pop-up menu listing these
functions will appear. Move the highlighted bar in the pop-up list
to the function you would like to add and press the <Enter>
key. This function will then appear in the EXEC pull-down menu and
can be used like any of the other image processing routines in EXEC.
Beyond these, you can examine individual pixel values in an image
through the "Pixels" pull-down menu. For more on this, see the
section "Examining Pixels."
ALL IN AN IMAGE
A computer image is made up of individual picture elements called
pixels. The images in EyeSight are restricted to 128 x 128 pixels
in size. A pixel represents the intensity of light in the image
at its location in the scene. For the images in EyeSight each pixel
can take on a value between 0 and 255, where 0 is black and 255
is white. The values inbetween are various shades of gray.
Note: EyeSight does not display all the gray-levels in an image.
For instance, on a CGA graphics card, EyeSight will only display
four gray-levels. For more on this, see the section entitled
"Image Colors."
WHAT YOU SEE ...
EyeSight provides two windows for displaying images - one on the left
side of the screen and the other on the right. In addition, the
left window allows you to keep multiple images in memory at one time.
THE LEFT DISPLAY WINDOW
The left window is where all image files are read into and holds the
input image used in all image processing function. Even though you
can only display one image at a time in the left window, EyeSight
allows you to have multiple images open at once by placing them
into a circular list of buffers. (The image currently being displayed
- Page 7 -
______________________________________________________________EyeSight
in the left window is often called the current image.) You can step
through this list, and consequently change which one is displayed,
with the F4 function key. Alternatively, you can get a list of all
the currently open image buffers by holding the ALT key and pressing
the letter b. From this list of buffers you can use the up and down
arrow keys to select one of the buffers and then press <Enter> to
have its image displayed in the left window.
THE RIGHT DISPLAY WINDOW
The right display window has only one image buffer. Its primary
purpose is for displaying the results of an image processing
operation. By placing the results in the right window, the original
image (the image in the left display window) will not be over-
written and will stay intact.
IMAGE COLORS
Each pixel in an image is represented by one pixel on the screen.
The color on the screen indicates the value of the pixel. Even
though each image pixel can range between 0 and 255, you will
not see all of the variations. For instance, on a CGA graphics
card you will only see one of four colors displayed on the screen for
any given pixel. The EGA mode also can only display four intensity
levels but it uses four gray colors to accomplish this. VGA, on
the other hand, can display 64 gray-levels.
On a CGA the colors that are used to display an image correspond
to the intensity values shown in the table below.
Color on the screen Pixel Value Range Actual intensity
================================================================
black 0 to 63 black
magenta 64 to 127 dark gray
cyan 128 to 191 light gray
white 192 to 255 white
To configure EyeSight to use a particular mode, see the
section SETTING THE GRAPHICS MODE.
SELECTING THE DISPLAY WINDOW
The output of all image processing operations is directed to the
window above the " Result ^" message. When the program begins,
"Result ^" defaults to the right window. You can alternate between
the two display windows by pressing the TAB key. If "Result ^" is on
the right side of the screen, (as it is by default) the result of
all image processing functions will be displayed on the right and
the original image will stay intact. If the result is to be placed
on the left side of the screen however, the original image will
be overwritten and lost.
- Page 8 -
______________________________________________________________EyeSight
Note: Even though you can control where the output of an image
processing function goes, all the vision processing functions use
the current image from the left side of the screen as input.
COPYING BETWEEN DISPLAY WINDOWS
In order to provide a flexible environment for processing and
displaying images, EyeSight allows you to copy images between the left
and right windows. A "Duplicate" option exists under the "Images"
menu. However, there are several quick keys which it informs you of
that you can use directly. These commands are:
Keys to use Description
============================================================
F5 Copy currently displayed left image
to the right display window
F6 Copy the image in the right display
window to the left image. The
currently displayed left image will
be over-written
SHIFT-F5 Copy the current image in the left
window into a NEW image buffer
(which this function will automat-
ically allocate) on the left side
of the screen. This effectively
creates a duplicate of the current
image
SHIFT-F6 Allocate a new image buffer in the
left window and copy the image in
the right window to it
These copy commands will work with all images, except the results
of the histogram, profile, contour, and Burns functions.
READING AN IMAGE FILE AND MULTIPLE IMAGE BUFFERS
As mentioned earlier, EyeSight supports multiple image buffers for the
left display window. As a convenience, when EyeSight begins it
automatically sets up a buffer for the first image to be read into.
Therefore, when the "Load Image" option is selected, the image will
be read directly from the image file to this buffer and displayed
on the left side of the screen. The "Load Image" function will prompt
you for the filename to read. If you use wildcard characters in
the filename EyeSight will display a list of files that match its
specifications from which you can select which file to be read.
By default, the "Load Image" function always reads an image file into
the current image buffer in the left display window. However, if an
image already exists in the current image buffer, EyeSight will prompt
you if its okay to overwrite the buffer with the new image.
- Page 9 -
_____________________________________________________________EyeSight
Alternatively, you can manually create a new buffer first by using
the "Open Image" selection in the "File" pull-down menu and then
reading the image file into this buffer by performing the "Load Image"
function. Yet another way to read an image into a new buffer is to
use the SHIFT-F2 key combination. This will invoke a function which
allocates a new image buffer and then reads an image into it.
(An alternative way of opening a new image buffer is to use the gray
'+' key. Once again, to swap between the multiple buffers use the F4
function key or the ALT-b command.)
You can close a buffer by selecting the "Close Image" command also
under the "File" menu bar. The grey '-' key performs the same
function and makes a handy way of quickly deleting image
buffers. Closing a buffer will erase the current image (in the left
display window), delete it from the list of images, and then display
the next image in the list, if one exists.
READING A FILE FROM THE COMMAND-LINE
You can also instruct EyeSight to read in an image file by specifying
its name when you invoke the program. For instance, to read in the
clips image when the program is invoked, you need only type:
eyesight clips.im <Enter>
If there are other images you want EyeSight to read, you may also put
them after the EyeSight name -- each image file name should be
separated by a space. The images will be placed into separate
image buffers in the left display window and the last image in the
list will be displayed in the left window.
LISTING THE IMAGE BUFFERS
You can pop-up a list of the currently open image buffers by
holding down the ALT key and pressing the letter b. You can
switch the current image to one of these buffers by using
the up and down arrow keys to select a buffer by highlighting
it and then pressing <Enter>. The image corresponding to the
selected buffer will appear in the left display window.
WRITING AN IMAGE FILE
You can write the current image to a disk file with the "Save Image"
option under the "File" selection in the main menu bar. The file
is stored as a binary file with a special header that is used
internally by EyeSight.
- Page 10 -
______________________________________________________________EyeSight
In addition to this, each buffer has a name associated with
it. This name is used as the filename for the "Save Image" command.
You can change the name of the image with the "Rename Image" option
under the "File" header.
DEFAULT PATHNAMES
EyeSight also provides a set of default path prefixes that are used
when reading or writing image files. This is particularly useful
when you want to store images in various directories. The path
prefixes are:
Pathname Description
=======================================================
input image Directory image files are read from.
output image Directory image files are written to.
current Directory used for internal dir.
DOS Directory holding COMMAND.COM
When EyeSight begins, the path prefixes are set to NULL so that all
image files default to the current directory.
Note: If you specify a directory prefix you should not place a "\"
symbol at the end of the pathname.
For most uses of EyeSight you will never need to alter these path
names. EyeSight will always default to using the current directory
which is adequate for most operations.
CONFIGURATION FILE
There are several user-definable constants in EyeSight - threshold
values, graphics mode, path prefixes, and so on. These can all be
defined in a configuration file. When EyeSight begins it looks
in the current directory for a text file called
eyesight.cfg
If this file exists, the specifications in the configuration are
used. If the file does not exist, the system defaults to the
internal configurations. Once in EyeSight, you can read in any
configuration file with the "Option", "Load Configuration" command.
In addition, you can save the current state of all settings in
EyeSight at any time to a configuration file, by using the
"Save Configuration" command under the "Option" pull-down menu.
- Page 11 -
______________________________________________________________EyeSight
HELP!
EyeSight provides two levels of help. Pressing the F1 function key
will initiate a general help system that is divided into several
topics. You can select topics from the help screen by using the
cursor keys and pressing <Enter> when the topic desired is
highlighted.
In addition, EyeSight provides a hypertext, context-sensitive
help that is accessible with the SHIFT-F1 key combination. It
provides short help descriptions about your current location in the
environment. For example, if the highlight bar is on the
"Threshold" option and SHIFT-F1 is pressed, a brief help message
will appear explaining the threshold operation. Words that appear
in color within a help window act as hypertext links to other
help displays. If you want help on such a topic, move the highlight
bar in the help screen (using the arrow keys) to the word and
press Enter. To exit a help screen, press ESC.
INTERNAL DIRECTORY
You can get a listing of the current directory (the directory
specified in the OPTIONS/ CURRENT DIRECTORY option) by selecting
the DIRECTORY option under the FILE option. You may use a wild
card specification to restrict the directory listing.
EXITING THE PROGRAM
You can exit EyeSight at almost any point in the program by typing
ALT-x. Alternatively, you can elect to suspend the program by
pressing ALT-z and thereby enter the DOS environment. This option is
handy if you have forgotten a filename and want to go out searching
for it. Once you have suspended the program you can return to
where you left off in EyeSight by typing "exit" at a DOS prompt.
PRINTING AN IMAGE
Another utility provided in EyeSight is a print screen function. You
will find it under the "Images" menu header. For this function you
will need an IBM compatible graphics printer connected to your PC.
This function will only work in CGA mode. The function uses the DOS
print screen call, so you must first initialize the systems to
graphics mode by executing GRAPHICS.COM which is included in your
DOS distribution disk. You only need to run this program once before
you run EyeSight and only if you are going to print the screen.
If you select this option, you will see the screen redrawn with
the negatives of the images, then the printing enabled, and finally
the screen restored to its original form. The image negation is
needed in order for the printing to appear as it does on the screen.
- Page 12 -
______________________________________________________________EyeSight
EXAMINING PIXELS
You can check the location of pixels in an image and their values
with the "pixel Values" option under "Images". This selection will
let you move a cursor to a particular image location and display its
coordinates and pixel value. The cursor defaults to the "Result ^"
window. You can toggle between the left and right images with the
TAB key. In addition, if the "Result ^" window is pointing to the
left display window, you can step through the image buffers using
the F4 function key. To exit the pixel inspection routine, use the
ESC key.
The arrow keys move the cursor around the image. You can toggle the
"Insert" key below the keypad to alter the amount that the cursor
moves by. At any time you can jump the cursor back to the center of
the image by pressing CTRL-HOME.
TWO IMAGE OPERATIONS
Some image operations, like add and subtract, require two images
as input. Therefore an image marking function is provided so
that you can select which images in the list of image buffers
will be used in the operation. All the functions that use
two images use the current image in the left display window and
the currently marked image as input.
To mark an image, first step through the list of images until
the desired image is in the left display window. Then hold down the
ALT key and press the letter m. This will mark that image. To show
that the image has been marked, its name will be shown in reverse
video. The marked image will be used in all succeeding two image
operations until you unmark it by pressing ALT-m again or when
you mark another image.
Once you have selected the second image, step back through the
list of images until you get to the other image you want in the
image operation. The result will be shown in the Result window
as before.
If there are only two images in the list of images, then you do
not need to mark either of the images -- it will automatically
use the two images.
IMAGE ANALYSIS FUNCTIONS
The following is a very brief overview of the image processing
functions supported in EyeSight.
- Page 13 -
______________________________________________________________EyeSight
Histogram - This function graphically displays the number of
occurrences of each intensity value in an image. All possible
gray scale values (intensity values) are plotted on the horizontal
axis and the number of occurrences of each of the gray scale values
is represented by the vertical axis. Histograms are useful for
determining good threshold values and checking the intensity
distribution in an image.
Threshold - Converts the current image to a black and white image.
All pixels below a threshold value are set to black and all other
values are set to white. Converting gray scale images to binary form
is quite common in industrial systems. It simplifies the processing
since the vision program only has to deal with two intensities --
black and white. Usually contour followers are used in conjunction
with thresholding operations to locate objects. See the description
of the contour operation.
Profile - This function graphically displays the pixel intensities
for a single line in an image. This routine can give you a good
idea how strong edges are in an image.
Roberts Gradient - Named after its developer, the Robert's gradient
is an edge detector known for its simplicity and speed. Edge
detectors try to locate edges of objects in an image in the hopes
of being able to identify them. Although the Robert's gradient
doesn't yield results that are as good as many other edge detectors,
it is acceptable for exploring computer vision. As with each of the
edge detectors in EyeSight, the roberts function will display a
gradient image where each pixel indicates how strong edges are at
each point in the image. The stronger the edge the greater the
image pixel value displayed in the gradient image.
Average - Reduces noise in an image by averaging small 3 by 3
neighborhoods of pixels. This operation will also tend to blur
an image.
Sharpen - Enhances the edges of images by subtracting the Laplacian
of the image with itself. Unfortunately, it often enhances noise
too. It's fun to experiment with this routine -- especially when
the image is a little blurred.
Sobel - One of the most common edge detectors. It detects
edges reasonably well although it does take a bit more time to
compute than say the Roberts gradient. (See Roberts.)
Prewitt - Yet another edge detector. It works much like the Sobel
edge detector. The main difference is how the pixel neighborhoods
are weighted when the edge operation is performed. (See Roberts.)
And - Performs a bitwise AND of two images.
- Page 14 -
______________________________________________________________EyeSight
Or - Performs a bitwise OR of two images.
Add - Adds two images together. All values are truncated to 255.
Invert - Inverts an image so that black becomes white, white
becomes black, and so on.
Subtract - Takes the absolute value of the difference between two
images.
Contrast - Improves the contrast of an image. It removes a
portion of the image's intensity values from the top and bottom
of the histogram. The rest of the image is stretched to fill
the full range of intensity values.
Log - Calculates the logarithm of an image and rescales the
resulting intensity values so that they range between 0 and 255.
Consequently, it alters the distribution of the gray scale values
in an image. Some researchers theorize that the human eye
automatically performs a log operation on all scenes so that
we can see greater detail in shadowy areas.
Burns - A line finding algorithm that uses regions of edge gradients
as a basis for locating lines. This algorithm is excellent for
detecting a wide variety of edges; however it is memory greedy and
computationally expensive. Nonetheless, finding lines can be
a useful step in higher level vision processing.
Contour - Thresholds the image and then divides the result into
black and white regions that can be used for object recognition.
The next section discusses it in greater detail.
THE CONTOUR FOLLOWER
Under the "Exec" menu selection is the Contour option. This routine
is an implementation of a Binary Region Tracer. Quite simply, the
routine traces the outlines of any regions it finds in a black and
white image (hence the name binary). Binary Region Tracers were
developed as the result of early vision research towards object
recognition. They have found greatest acceptance on the
industrial assembly line where the lighting can be strictly
controlled.
The routine implemented in EyeSight accumulates statistics
about each region's area, perimeter, color, and so on as it
scans the image. You can inspect this data after the contour
routine has been run by pressing ALT-d. After doing this, one
of the regions will be marked with a cursor and a window will
appear on the other side of the screen with its statistics. To
inspect the statistics for other regions, use the up and down
arrow keys. Press ESC when you are finished.
- Page 15 -
______________________________________________________________EyeSight
Since the routine will only work on a black and white image, each
time you call the contour program it will prompt you for a
threshold value to apply against the image. The contour result will
be displayed in the result window.
To fully understand the limitations of the contour routine, try
various threshold values on each image and see what happens.
EXPERIMENTING WITH EYESIGHT
This section provides a hands-on tour through some of the
facilities included in EyeSight. You should already have read
the previous sections in this manual before continuing with
this section.
In particular, we will use EyeSight to demonstrate some of
the differences between the Roberts and Sobel edge detectors.
The following set of step-by-step instructions will guide
you through reading an image file, calculating its gradient
using the Roberts and Sobel edge detectors, and then subtracting
these two intermediate results to illustrate the differences
between them. For this example we will use the clips.im image
file. You may want to experiment with others later.
Step 1: Begin the EyeSight program, if you have not already
done so, by typing
eyesight
at the DOS prompt and pressing the Enter key. EyeSight will
appear with an author message, simply press Enter to continue.
Step 2: Read the clips.im image file. This is done by moving the
highlight bar to the File pull-down menu using the left and
right arrow keys (if it is not already selected) and then
selecting the Load Image function in this pull-down menu by
moving its highlight bar to Load Image and then pressing
Enter. The program will then prompt you for an image file
name. You should now type:
clips.im
and press Enter. The image of a set of black clips on a white
background will appear.
Step 3: Now let's calculate the Roberts gradient for this image.
The Roberts gradient is located in the Exec pull-down menu. To
get to it, move the highlight bar to Exec using the left and
- Page 16 -
______________________________________________________________EyeSight
right arrow keys and to the word Roberts within its pull-down
menu using the up and down arrow keys. Press Enter to execute
the Roberts function.
On the right of the screen will appear a gradient image of
the original clips image. It indicates how strong edges are
at each pixel in the image. The stronger the edge, the higher
the pixel value (and brighter the color used to display the pixel).
Step 4: We will be using the Roberts gradient image later so
we need to copy it to our list of image buffers on the left
side of the screen. To do this, hold down the Shift key and
press the F6 function key. The Roberts result will now appear
on the left side of the screen. (Note: the original clips
image is still intact since we have copied the gradient
image into a NEW buffer.) It is given a name of NONAME.IM.
Step 5: Give the gradient image a name. Since we will be
referring to the Roberts gradient result later, let's give it
a new name. This is accomplished using the Rename Image function
contained in the File pull-down menu. To use it press ALT-f (or
use the arrow keys to get to the File pull-down menu), move the
highlight bar to "Rename Image" and then press Enter. EyeSight
will prompt you for the new image name. Type in:
clips.rob
and press Enter. The name of the gradient image is now clips.rob.
Step 6: Now let's get back to the original image of the clips by
pressing the F4 function key. The clips.im will now appear.
Step 7: Perform the Sobel edge detector on the clips image.
Before we can do this we must add the Sobel function to the
Exec pull-down menu. This is done by selecting the Add Function
operation under the Option header in the main menu. To do this
press ALT-o, followed by the letter 'a'. (Alternatively you can
select this function by using the arrow keys.) A pop-up window
listing several functions will appear in the middle of the screen.
Move the highlight bar in this window down (using the down arrow
key) until it is on the word sobEl. Press the Enter key. Now the
Sobel function is available to you in the Exec pull-down menu. To
use it press ALT-e and move the highlight bar until sobEl is
selected and then press Enter.
This will execute the Sobel function. (Note that the Sobel
function takes longer to execute than the Roberts gradient did.)
Step 8: Copy the Sobel result to the image buffer list on the left
side of the screen. This is done by holding down the Shift key and
pressing the function key F6. (The Sobel result will now appear in
the left window.)
- Page 17 -
______________________________________________________________EyeSight
Step 9: Although it is not necessary, let's give the Sobel
gradient image a new name. This is accomplished, as before,
by using the Rename Image function in the File pull-down
menu. To do this type ALT-f followed by the letter 'r'.
Then type:
clips.sob
at the prompt.
Step 10: Now to subtract the two gradient images... First
we need access to the image subtraction function in EyeSight.
It is also included in the Add Function list under the Options
menu. To return to this list press ALT-o and then the letter 'a'.
Move the highlight bar in the pop-up list of functions, that
should now appear, to the word sUbtract and press Enter.
The subtraction routine should now be in the Exec pull-down
menu.
Step 11: When we execute the image subtraction routine it will
use as input the current image (shown in the left display window)
and a marked image. We will select and mark an image by pressing
the ALt-m key combination when the desired image to mark is in
the left display window. In this case we will make clips.sob the
marked image. Since it is already the current image, simply
type ALT-m. The name clips.sob should now appear in bold to
indicate that it is a marked image.
Step 12: Now move to the clips.rob image in the circular list
of image buffers by pressing the F4 function key until it appears.
You should only have to press F4 twice.
Step 13: And finally (!) let's subtract clips.rob and clips.sob to
illustrate the difference between the Roberts and Sobel edge detectors.
To do this, press ALT-e, move the highlight bar to the word sUbtract,
and then press Enter. The difference between the two edge gradient
images will now appear on the right side of the screen.
In practical terms this example illustrates that not all
edge detectors are made equal. As you can see, the Sobel edge
detector has done a better job of detecting object edges -- although
it was a little slower to do so.
To see how significant the differences are, you might want to
try using the pixel Values function in the Images pull-down
menu or run the histogram on the difference image (To do the latter
you should first copy the difference image to the left display
window, since all image functions operate on the current image in
the left display window.)
- Page 18 -
______________________________________________________________EyeSight
If you've made it this far, you are well on your way to using
EyeSight to its fullest and will probably be discovering (or
rediscovering) just how amazing our own vision capabilities
really are! Good-luck!
SCRIPT LANGUAGE
EyeSight is going through a change. EyeSight is being equipped
with a powerful script language that provides access to each of
its functions. Unfortunately, this language interface is not
complete so it is not presented here. However, the EyeSight
distribution disk does contain a short script file, demo.scr,
that demonstrates a handful of the EyeSight language primitives.
To run this demonstration file type at the DOS prompt:
eyesight demo.scr
This script file will execute a few of the functions provided
in EyeSight and will display brief statements on each. You
may want to take a closer look at this file.
SETTING THE GRAPHICS MODE
EyeSight supports three video adapters: the CGA, EGA, and VGA.
The CGA is used by default. (For more information on what each
of these modes can display, see the section entitled: Image Colors.)
You can select which graphics mode to use by specifying in the
configuration file the type of adapter installed. This is done
with the "DISPLAY" command. For example, to configure EyeSight
to use CGA, the command is:
DISPLAY = CGA
Similarly, to use EGA:
DISPLAY = EGA4
and finally, to use VGA, use:
DISPLAY = VGA64
Three sample configuration files are included with the distribution
disk that already have these values in them. The one CGA.CFG is
for the CGA, EGA.CFG for EGA, and VGA.CFG for the VGA. If you
want to use a mode other than CGA (remember, EyeSight uses CGA by
default), you should rename the file corresponding to the mode
you'd like to use to "eyesight.cfg" and place this file in
EyeSight's directory. Then when you start EyeSight it will appear
in the mode selected.
- Page 19 -
______________________________________________________________EyeSight
Tip: The EGA mode is the slowest, so if you have an EGA card, you
may want to use the CGA mode anyway-- the difference in speed will
more than make up for the inability to display images in gray
colors.
Warning: On some VGA cards, the CGA colors will not appear correctly.
For these boards, you should only use the EGA4 or VGA64 settings.
COMMAND SUMMARY
Although all of the commands can be accessed through the cursor
keys, they may also be selected by entering one or two letter
command sequences and some even have specially defined quick-keys.
The following is a list of all of the commands available
in the default configuration of EyeSight.
These commands do not appear in the menu system:
Key Sequence Description
===================================================================
ALT-b List all image buffers -- enables you to
jump to another image buffer by selecting
a buffer from the list and pressing <Enter>
ALT-m Mark an image to be used in a two operand
image operation like and, subtract, or add
The following commands are listed in the menu system. You can
use a combination of the arrow keys and the <Enter> key to select
one of these functions or you can elect to use the key sequences
shown.
Key Sequence Description
===================================================================
ALT-f, l load an image file into the current buffer
ALT-f, s save the current image buffer to a file
ALT-f, o open a new image buffer
ALT-f, c close the current image buffer
ALT-f, n display the next image buffer
ALT-f, r rename the current image buffer
ALT-f, d list files in the current directory
ALT-e, h display histogram of current image buffer
ALT-e, p display a profile of the current image buffer
ALT-e, t threshold the current image
ALT-e, r calculate and display roberts gradient
ALT-e, a average the current image
ALT-e, c trace the contours of a b&w version of the image
ALT-e, b find lines in the current image
- Page 20 -
______________________________________________________________EyeSight
Key Sequence Description
===================================================================
ALT-s take a picture from a camera - NOT SUPPORTED
IN EyeSight
ALT-i, v examine pixel values in the Result window
ALT-i, p print the screen
ALT-i, c clear the display window above the "Result ^"
message
ALT-i, t toggle the Result window to the other display
window
ALT-i, d make a duplicate of an image
ALT-o, p change the default directory prefixes
ALT-o, w save the settings of the program in a
configuration file
ALT-o, l load a configuration file
ALT-o, a add a function to the "Exec" pull-down menu
ALT-o, d remove a function from the "Exec" pull-down menu
ALT-o, s suspend the program and return to DOS
Beyond the standard commands there are several quick access commands.
These are:
QUICK COMMANDS:
Key Sequence Description
===================================================================
TAB change window for vision processing results
This window is called the result window.
ALT-c clear Result display window
ALT-p perform a print screen
ALT-x exit program
ALT-z suspend program, exit to DOS
+ open an image buffer
- close an image buffer
F1 help
F2 load an image file into the current image buffer
F3 write the current image buffer to a file
F4 display next image buffer
F5 copy current image buffer to right window
F6 copy right image to current image buffer
SHIFT-F1 quick command reference sheet
SHIFT-F2 load an image file into a NEW buffer
SHIFT-F5 copy current image into a new image buffer
and display it in left window
SHIFT-F6 copy right image into a NEW buffer and
display it on the left side of the screen
- Page 21 -
VISION.DOC
A Quick Introduction to Computer Vision
Loren Heiny
Robots, Etc
P.O. Box 122
Tempe, AZ 85280
(A version of this article first appeared in the Winter
1989 issue of Robot Review, Vol 2, No. 1, pp. 1-5.,
950 Trout Brook Drive, West Hartford, CT 06119.)
Vision is amazing. When you sit back and think about how
much detail and information is almost instantly available to
us as we gaze around, you begin to appreciate the real
complexity and amazing power of our vision system.
As a robot experimenter, I look forward to the day when
my robot can see what I can. Unfortunately, we are far from
this goal. Part of the problem is the lightning fast
computing speed that is needed, but what it really comes
down to is that we simply don't know yet how to build such a
vision system.
Despite this, researchers have developed volumes of
computer vision algorithms, ideas, and strategies. Some are
even useful. In this article we'll take a quick tour of two
computer vision techniques which are often used to implement
vision systems that recognize objects -- region analysis and
edge detection. These two techniques represent only a small
portion of the vision field, but they will introduce you to
some of the basic concepts involved in computer vision so that
you can begin thinking about a vision system for your own
computer or robot.
What's in an Image?
A computer image is made up of (usually) tens of thousands
of tiny elements, called pixels which are much like the tiny
dots that make up the picture on your TV set. In computer
vision analysis, it is up to the computer to scan through
these thousands of pixels and figure out *what objects are where*.
Unfortunately, each pixel taken by itself, does not
really provide much information. In fact, all it does
indicate is the average intensity of light at its location
in the scene. Converting these intensity values into real-
world objects is no easy chore!
Object Recognition
Recognizing objects is an extremely important part of
computer vision. There are two techniques commonly used --
region analysis and edge detection. Although both are used
in recognizing objects, they each go about it by from two
seemingly opposite directions. Region analysis, on the one
hand, attempts to recognize objects by grouping regions of
like pixels and then trying to match these regions with the
shapes and sizes of known objects. In contrast, edge
detection attempts to recognize objects by locating their
edges by looking for unlike pixels and then basing its
recognition upon the contours of these edge points. Neither
technique is fool proof, but each has found acceptance in
many areas. Let's begin by looking at region analysis more
closely.
Region Analysis
Region analysis, in principle, is very simple -- regions of
like pixels are grouped together to form objects. Usually,
an image is scanned from top to bottom, comparing each pixel
intensity with its neighbors. If adjacent pixels are
similar, then they are assumed to belong to the same object,
otherwise the pixels are assumed to be located on different
objects or part of the background.
Probably the most successful form of region analysis
is called binary region analysis. Here, images of objects
laying on a highly contrasting background are first
converted so that they only contain two intensity values (or
gray-levels as they are more commonly called) -- black and
white. This process is called thresholding because all
intensity values above a particular value, the threshold,
are set to one gray level (white, for instance) and all
others to a second value (in this case, black). As long as
all objects contrast with their background, they will
become one intensity and the background the other.
Once a binary image has been generated it is a matter of
grouping all connected pixels that are similar into regions
so that they can be matched with known objects. The matching
process is usually based on such things as the area,
"color," perimeter, and number of holes of the regions.
To make the technique work well, however, several
conditions usually must be met. First, there must be good
contrast between the objects and the background. Second, in
order to maintain this contrast, many times the lighting
must be known or at least controlled. And thirdly,
recognition of objects usually requires that they are always
viewed from the same orientation and perspective.
Although, each of these conditions can become very
restrictive, they aren't always a problem. For instance, on
assembly lines with overhead cameras and controlled
lighting binary region analysis can be very reliable and
useful.
One of the reasons binary imagery is used is because it
is simple to implement. Therefore, it is worthwhile to try
to find some uses for it in our household robots. One
possibility, is to use binary region analysis to recognize
highly contrasting objects on a wall. For instance, your
robot might be able to recognize a dark door handle on a
white door by recognizing its circular contour and general
height above the floor. Another possibility is to "read"
special black-and-white markers that you have placed around
the house so your robot can see where it is.
--------------------------------------------------------------
EyeSight includes a binary region analysis function, called
Contour, which is located in the Exec pull-down menu. To run
this routine select the Contour command as outlined in the
EyeSight manual. It will display a slide-bar from which you
can select a threshold value which will be used to convert
the current image to binary form. (Use the left and right
arrow keys to select a threshold value and then press the
Enter key to accept it and begin processing.) Once the
threshold is chosen the binary version of the image will be
displayed followed by the outlines of the regions that were
found in the image.
You can examine several of the statistics accumulated
on each region by pressing the ALT-D key combination. This
will cause a pop-up window to appear with the region
information for the region currently marked by a crosshair.
Use the up and down arrow keys to sequence through the
various regions in the image. Press ESC when you are done
and want to continue with EyeSight.
--------------------------------------------------------------
Edge Detection
There are many computer vision programs that base
their analysis not on the regions found in an image but
rather its edges. The basic idea is to scan the image
looking for significant changes in neighboring pixels.
This approach assumes that changes in adjacent pixels
will occur at the edges of objects. Therefore, once these
edge points are located, they can be chained together
and used to recognize what object they come from. Once
again, this is no easy chore.
First, it is generally not easy to locate all of the
object edges in an image. In addition, it turns out that
many image edge points may not correspond to actual object
edges at all. For instance, shadows, texture, and lighting
variations can all create strong edges in an image that
obviously do not correspond to any real object edges.
Secondly, it is often very difficult to link edge points
correctly. Many times not all the edge points of an
object are strong enough to be detected. Therefore the
contour of the object is broken. Consequently, it is often
necessary to have to link small chains of non-connecting
edge points together to produce a complete chain of the
object's contour. This can be difficult because it is often
a challenge to decide if two chains should really be
connected or not.
-----------------------------------------------------------
EyeSight:
EyeSight includes three edge detectors, each which is
named after its developer: Roberts, Sobel, and Prewitt.
Each edge detector uses a slightly different method to compare
neighboring pixels, although they all generate an edge
gradient image which indicates how strong edges are at
each location in an image.
The Roberts edge detector is one of EyeSight's standard
functions and appears under the Exec pull-down menu. The Sobel
and Prewitt edge detectors, however, must be added to the
Exec menu before they can be used. This is accomplished by
selecting them from the Add Function command contained in
the Option menu. (Refer to the EyeSight manual for more
information on how to do this.)
Each of the edge detectors, when selected, will display
a gradient image for the current image in EyeSight as
described in the EyeSight manual. In general, the stronger
the edges are in the gradient image, the brighter the pixels
will be. You may want to compare the results of each of the edge
detectors as described in the section EXPERIMENTING WITH EYESIGHT
in the EyeSight manual.
-----------------------------------------------------------
Line Finding
One common extension to edge detection is to combine
chains of edge points into lines and then base the image
analysis on these lines. This process, called line finding,
is quite attractive because it often simplifies object
recognition since there are far fewer lines in an image to
match or identify than edge points. In addition, line
finding often leads to fewer mis-matches and faster
processing.
-------------------------------------------------------------
EyeSight:
EyeSight includes the Burns line finding algorithm that
finds lines in an image by grouping regions of like *edge*
pixels and fitting lines to these regions. The Burns function
is located under the Exec pull-down menu. To find lines in
an image all you need to do is select the Burns command
(assuming you've already read in an image to the EyeSight
environment). Try experimenting with several different images
and note the number and quality of the lines found in each.
Alternatively, try guessing what is in the original image
based on the lines you see on the screen. This can be a
real challenge!
----------------------------------------------------------------
Other Techniques and Closing Comments
So far we have briefly discussed some of the issues involved
in object recognition, but there are many more computer
vision techniques that have been developed over the years.
For example, researchers have used texture, shading, color,
focus, modeling, multiple image analysis (motion and
stereo), light striping, and much, much more to attempt to
interpret images.
However, despite all of this effort, no one can yet
claim to have a vision system that can even remotely rival
our own. This is still a challenge yet to be met. Of course,
it does need to be said that a robot (or your computer) does
not necessarily have to be able to see the same way we do.
For example, it's probably not desirable to intentionally
program our robots to experience the same optical illusions
that we do--or is this unavoidable in a vision system that
equals our own?
Directory of PC-SIG Library Disk #1678
Volume in drive A has no label
Directory of A:\
BALLOON IM 16512 3-11-89 3:06p
BOLTS IM 16512 3-11-89 2:58p
CGA CFG 189 3-13-89 10:37a
CLIPS IM 16512 3-11-89 3:03p
COMPUTER IM 16512 3-11-89 3:04p
DEMO SCR 597 1-15-89 3:39p
EGA CFG 190 3-13-89 10:44a
EYESIGHT EXE 154979 4-12-89 8:07p
EYESIGHT HLP 22096 3-13-89 9:12a
FILE1678 TXT 2443 12-07-89 12:31p
FLOWER IM 16512 3-11-89 3:07p
GO BAT 38 1-01-80 1:37a
GO TXT 540 12-08-89 12:47p
MANUAL DOC 58105 4-12-89 11:33a
MTN IM 16512 3-11-89 3:05p
VGA CFG 191 3-13-89 10:44a
VISION DOC 11668 4-07-89 1:12p
17 file(s) 350108 bytes
0 bytes free