CS 4495/7495: Computer Vision
Problem Set 1 (Fun with Perspective and Matlab)
http://www.cc.gatech.edu/classes/AY2002/cs4495_fall/ps/PS1.html
Handed Out: Aug. 23, 2001
Due: August 30, 2001 (before class)
The main aim of this Problem Set (PS) is to familiarize you with the idea of perspective and to gain some experience with Matlab. Some of you are already experts on Matlab. If you are, help your classmates. Also, for the experts, please pay attention to the bottom optional part of this PS as it is only for you. We will also learn to use some other tools for this PS.
PART I:
Do Problems 2-1, 2-2, and 2-3 from the handout, Horn's "Robot Vision", Chapter 2, "Image Formation and Image Sensing."
Try to do these problems algebraically since we've already shown some of the answers (esp. 2-2) geometrically in class.
PART II: MATLAB fun
If you’re not using MATLAB do the equivalent on your system….
Convert the "oops" image below
to a MATLAB readable form. It
is both in your browser and also located at
/net/www/classes/AY2002/cs4495_fall/images/oops.gif (and maybe others)..
Matlab can take images in TIFF, JPEG, X, or RAW format. So the best thing to do is to convert images to one of these formats. The program "xv" may be of help.
MATLAB
1. Start matlab (typing "matlab" on a shell should do it).
o Play around for a while.
§ Try to understand the significance of variables.
o Type "demos" Note the Image Processing Toolbox
o Type imagedemo
o Type roidemo
o Play around as much as possible.
o Type "more on" This causes matlab to wait until you press a space before scrolling the next line of text up the screen.
o Type "help images"
o Commands to remember: "help", "help matlab/general", "lookfor"
o Type doc in matlab or matlabdoc on unix
o Check out the mathworks web page www.mathworks.com
2. Load your image using "imread" (see help imread). Read in [R G B] values OR [X, map]
3. Convert image to a grayscale image (hint: "lookfor gray") (You can also do this using UNIX utilities)
4. Figure out how to crop this image. Crop it so as to perfectly frame the luggage box.
5. Experiment with the "Demos" section by typing "help images" and find functions that interest you.
6. Try them out on your image
7. Figure out how to save these images (hint: "lookfor write")
The orignal image should be "oops.jpg", while the resulting few
images should be named "oops1.jpg", "oops2.jpg", ... etc.
PART III: Capture an Image:
SGI/Indy with an IndyCam (in 130 CoC)
· Use the command "capture". It will open up a small window with output from the indycam and a trivial user interface. It is generally faster to save images to /tmp (local disk), but the SGI will remove these files automatically when it reboots.
· If there is nothing appearing on the window, check for trivial things like the lens cap. The SGI command vcp (for "video control panel") can help with more controls and options on what signal to use.
SGI/Indy (in CoC clusters)
· The capture program on an SGI Indy saves a color image in SGI RGB format. To convert it to JPEG, use the program called "imgview" or "imgworks".
· Using imgview save the image as a JPG file.
Any other system: If you know of any other way of capturing images, more power to you.
The aim here is to digitize good quality
color and gray scale images, including one of your face well framed in the
image. For an example of what we want to be turned in, see http:/www.cc.gatech.edu/classes/AY2001/cs7495_fall/participants/thad/ps1/index.html.
On the CoC file system this file is located at /net/www/classes/AY2001/cs7495_fall/participants/thad/ps1/index.html.
Acquire other images, either from the WWW or
using the camera.