views:

78

answers:

2

Hello:

I would like to do an interesting project for a computer graphics course. I know that there is a lot of literature out there (i.e. SIGGRAPH conference papers). I have a very large range of interest with regard to computer graphics (i.e. image processing, 3D modeling, rendering, animation). However, I've only taken computer vision/graphics for 2 semesters and thus don't have too much background experience, except for the class projects that I had to do.

I've been looking through SIGGRAPH papers trying to see if there is anything that will be of interest to me but the literature is extremely vast. I was wondering if anyone has any topic suggestions, anything interesting that you ran across that you could recommend. I would prefer to do something fun yet slightly challenging (not really interested in making a shooter game).

If this question does not belong here, I apologize and please let me know where I should move it.

Thanks!

+1  A: 

Real time hand(s) detector.

You'll have plenty of interesting and fun applications with this.

zaf
Aha, for example you can build an augmented reality system upon it. Look here: http://sexdrugsandappliedscience.com/blog/2010/1/20/localization-and-mapping.html You can implement some weapon simulators, like holding a sword or a bow. This one requires a glove: http://www.youtube.com/watch?v=VnG_9yvEDU8 , the paper: http://www.cs.washington.edu/homes/jovan/papers/wang-2009-rth.pdf
overrider
+1 for the great links and the paper - never seen that before!
zaf
+2  A: 

Image Drawing animator. (the name is kind of misleading, but I didn't care much about it)

Anyway, the software does the following:

  1. Takes an image say a JPEG or BMP as input.
  2. Extract the lines from the image. (I used Matlab and Laplace transformations)
  3. Convert the static lines to Vector paths.
  4. Simulate drawing the image using the extracted paths.

In summary, you should give an image, for example a city scape, the program extract all lines and start drawing the buildings, streets and sunset lines, then finally add the colors one by one, until the full image is done.

medopal
I'm not sure it would look great, because automatically detected lines tend to differ from the usual notion of object borders (for example, consider shadows). But the area of non-photorealistic image based rendering is really cool (one could emulate different painting techniques), and the results are often impressive. For example see http://compression.ru/video/cartoonizer/index_en.html
overrider
this IS a pretty cool idea =) thanks.
Myx
@overrider, actually my application was doing something exactly like "Adriana Sklenarikova" photo from your link, then after drawing the lines, it starts shading the colors one by one, it was an old project, unfortunately i don't have a clip during its work.
medopal