Computational Neuroscience 2001b 0368.3013.01
Frequently Asked Questions
exercise
2.
1. Where can
I find the article: "A computational theory of human stereo vision"
by D.Marr & T. Paggio ? is it available online ?
A. At the library.
I don't know of an online copy of it.
2. I think there is
something wrong with the second link in Hezy's on-line papers page (http://www.math.tau.ac.il/~hezy/acoustic.pdf).
A. Try the mirror
(from http://www.math.tau.ac.il/~odedsc/cns01b/cns01b.html).
3. Is there another
source for David Marr's algorithm ?
A. The actual algorithm
could be retrieved and understood from David Marr's book "Vision"
4. What should
our matlab program do - just create the disparity map or perform edge detection
on the input as well?
A. You should test
the algorithms both with and without performing the edge detection.
5. What exactly
is the DOG filter?
A. Difference of Gaussian
("The mexican hat").
6. What is the
maximum size of the input images? Can we assume that it is 1000x1000 (maximum)?
A. Yes.
7. We based our
1st part on the "Cepstral Filtering" article. The authors admit that the
performance of the presented algorithm on natural images is much worse
than on RDS (as we have seen when running our implementation). Would this
fact be taken into consideration?
A. This is indeed
on of the parameters you should consider and discuss when comparing the
algorithms.
8. What is a disparity
map?
A. A disparity map
is a matrix containing the disparity for each pixel of the original images.
9. May we choose implement
an algorithm which contains features that weren't studied in class ?
A. Of course, but
make sure you have a good understanding of it.
10. We were wondering
if we can assume the following regarding the input pictures in the first
question (of the second project):
a. Pictures have been
taken from the same height, that is, all disparities will only have an
X element (and no "diagonal" disparities).
b. The Axes of the
two pictures will be parallel, that is, the X axis of the first image is
parallel to the X axis of the second image, and the same goes for the Y
axis.
A. Yes for both.
11. Regarding the first
question of the project: will the project be tested with files we provide
or on other files?
A. On other files.
But you are encouraged to add some running results on the document, and
add the input pictures to the projects directory.
12. Is the resolution
of the disparity map to be the same as the original pictures ?
A. Not necessarily.
13. Can you give us
some links for searching journals ?
A. You have a few
at : http://www.wisdom.weizmann.ac.il/~yoad/
14. Should we attach
a copy of the 2nd paper we used (not one of the two mentioned in the definition)
?
A. Please do.
15. Can we invent an
algorithm of ourselves instead of using a 2nd paper ?
A. Yes, as long as
it is a good one.
16. Is there an example
for the 2nd part (auditory) ?
A. Yes: here.
17. What format should
we use to save the disparity matrix ?
A. Bitmap (.bmp).
18. I found an article
that applies a method to calculate a disparity map,
but the explanation
of it's relevancy to the human vision is somewhat
shady and not elaborated
enough. Could I add my own justifications for the validity of such
a model ?
A. Yes. If you think
it is not relevant (at least in some aspects) you may note that too.
19. What type of file
should the project work on? Is it gray scale with color map of length
256, starting from [0 0 0] and ending with [1 1 1] in a linear way?
A. The input is a
gray scale bitmap file. You should retrieve the color map or any other
detail from there.
20. In the link
you gave us for stereo pictures, the disparity gets larger as the object
gets further. It is inconsistent with the eyes as far as I understand.
Are these "valid" pictures?
A. Put a finger about
20 cm in front of you, concentrate on it, an close you left eye and then
and your right eye. Note what happens to the background. Now do the same
when you fixate on the background instead of fixating on your finger. What
is the different ? For further details, look for 'disparity' in Marr's
book.
21. Hezy states in
his article that a 16000 x 16000 pixel (conventional constant resolution)
must be used to achieve the desired results. You said the maximum
size will be 1000 x 1000. How does the 2 combine?
A. Your program will
be tested using a maximum 1000 x 1000 bitmaps. However you may use any
pictures to check it yourself, and add the results to your report.
22. In ex 1 we are
supposed to produce gif files, but (and I quote): MATLAB does not produce
GIF files (due to patent
restrictions) what
format should we use?
A. The output file
should be BITMAP (.bmp) and NOT (as written in the project's description)
gif.
23. We are supposed
to produce one output file disparity.bmp. What about other images (during
the process; with various parameters; with/without edge detection)?
A. You may include
any image you want to discuss or demonstrate in the documentation. However
the only output required for the run test is the disparity map.
24. Concerning the
cepstral filtering paper: What should be the assumptions about the input
resolution? How far are the images? In the paper the writers assumed that
the images are in a constant distance, and 6-12 minutes correspond to about
20 pixels. But that requires large input images to produce good results
(larger than 1000X1000).
A. You cannot assume
the distance of the image - note that stereo vision works in various distances.
However you may assume the angle per pixels assumed in the article.
25. Does it matter
in what color map we save the image?
A. Please use the
'gray' colormap (choose the parameter to show best your expected results).
26. DoG filter - different
kinds of pictures "need" slightly different parameters of the gaussian
to give nice edges. Is it possible that the edge detection will be checked
after trying a few different parameters ?
A. This may be right.
But you can do so only if you think this way is a good model.
27. If I add examples
of running tests to the documentation - will these images count in deducing
points from the grade?.
A. You are encouraged
to add relevant pictures of your result, but within the page limits.
28. If we use an algorithm
that outputs both positive and negative disparities; Should we omit the
sign (use the same color for +a and -a ) or shift the disparity so we don't
have negative disparities ?
A. Shift it.
29. You wrote the output
should be in a BMP format . The input files , as far as I understand -
stay 'gif'?
A. No. It is BMP as
well.
30. May we assume
we are not suppose to recognize disparity larger than 10 pixels?
A. Yes.
31. ( On "Columnar
Cepstral..."): When do they use a 1 dimensional Fourier transform and when
a 2 dimensional one?
A. The examples (images)
imply a 2D FFT, but the appendix use 1D. The algorithm itself calls for
a 2D FFT, since it
is performed on an image. If only an horizontal shift is expected (or allowed),
it will mean that the peak should be looked for only on the X axis of the
cepstrum - but it is a 2D FFT in any case. The reason a 1D FFT have been
used in the appendix is just for the sake of the simplicity of
the explanation,
32. Regarding the CNS
project, should we scale the disparity map to produce a more comprehensible
picture or should the values in the disparity map represent the actual
disparity values (and thus be values between 1 and 10 which look all black
to dark gray)?
A. The output bitmap
file should be visually checkable; that is - the colormap you choose
should make relevant disparity visible.
33. I have some question
about a matlab function. Who may I ask ?
A. You may ask the
system for the phone of the Matlab advisor.
34. Are we suppose
to output the DoG as well ?
A. No. But you are
to implement it as written in the project description. When using edge
detector in the process of finding the disparity map you are to use the
DoG you created, and NOT any ready made edge detector.