CS 4495/7495: Computer Vision

Problem Set 5

Due: Tuesday November 13, 2001

---

1. Implement a hierarchical version of the Lucas and Kanade motion estimation. A very algorithmic description is given in the “Frame Rate Compression” paper by Bergen et al. Apply your method to the following two images:

Image 1	Image 2	Image 3

To help you show your results there is a file called displayFlow.m that is a Matlab file for showing vector images using arrow diagrams.

The motion algorithms that you have seen in class and in the papers you read are based on two fundamental assumptions:

a) For any patch in frame 1 there exists a patch in frame 2 with the same brightness values (constant intensity assumption)

b) The location of the patch in the two frames are close, so linear approximations can be used (linearization assumption).

In the next part of this assignment you will first generate some simple cases for which this is true, and
then some cases in which it is false. The file texSquare.m can generate textured or bland squares over textured or balnd backgrounds.

2. Generate two frames of a textured square over nontextured background
(set textureFlag=2, isize=64, vx=2, vy=2, noiseLevel=0).
Note that in this case any image patch in frame 1 has a corresponding patch in frame 2.

Run your algorithm on these frames. You should get approximately right displacements for the square.

Now run your algorithm without the coarse to fine. Do you still get

approximately the right answer? What assumption is violated ?

3. Generate two frames of a textured square over textured background
(set textureFlag=1, isize=64, vx=2, vy=2, noiseLevel=0).
Note that not all patches have a corresponding patch in the other frame. Which patches are these?

Run your algorithm on this sequence. Where does it give approximately the right answer, and where doesn't it?
What happens when you increase the window size?

4. Generate two frames of a nontextured square over nontextured background
(set textureFlag=3, isize=64, vx=2, vy=2, noiseLevel=0).
Note that some patches have a unique corresponding patch in the other
frame, while others have several corresponding patches. Which are these?

Run your algorithm. Where does it get the right answer and where doesn't it? What happens when you increase the window size?

---

WHAT to TURN IN?

This problem set requires a little bit more thinking and algorithm construction than the previous problem sets. Therefore it is essential that you turn in the following in your README/index.html file (You will not get full credit otherwise):

Include a brief description of how your program works. Be sure to discuss any differences between the algorithm given in the paper and your implementation. Also make sure you answer all questions listed at the end of each subproblem.

You will submit your work electronically (no paper). Instead, you will submit an HTML file. Create your own web page in the spirit of the template provided for PS2. The format of the page is up to you. Feel free to add as many result images, intermediate results images, and result images for different values of the control parameters as you think is necessary to demonstrate your work. However, the page should contain at least the following:

1. your name on top of the page,
2. links to ALL of your matlab code,
3. links to your result images converted to JPG format.

Your matlab code should be reasonably commented if you want us to understand it and give you credit/comments. Also include test scripts for each of the four subproblems: test_1.m, test_2.m, test_3.m, and test_4.m. These files should demonstrate your work for parts 1,2,3, and 4 respectively. The scripts should:

1. load your test images,
2. call your other matlab functions that do the required image processing,
3. show all results on the screen using multiple figures, and
4. save all result images in JPG format with appropriate names
that can be loaded from the web page.

To submit your work do the following:

1. On a UNIX platform create a directory named after your CoC login_name
2. Put ALL your matlab files, the three test scripts, the original images, and all result images in that directory.
3. Also put a README file that explains any special needs of your code, and lists any KNOWN BUGS/FEATURES.
   For example if your code is running very slow try to explain why. If something is not working explain why.
   Also remember to put the required explanations of your code as described above.
4. Create an HTML page according to the instructions given above. Save it as index.html in the same directory.
5. CD to the parent directory.
6. Create a tar archive using:
   
   tar cvf login_name.tar login_name
7. Gzip the tar file using:
   
   gzip login_name.tar
8. E-mail the login_name.tar.gz file to cs7495@cc.gatech.edu

---