4

(A newbie in computer vision)

Goal is to reconstitute a game level using image stitching or any other method. The level played by someone is video-recorded, these frames will be the input.

Expected result : (level 4-4 of SMB from http://www.vgmaps.com/) enter image description here

This is my first attempt at addressing this problem, using OpenCV (EmguCV). So far results are excellent but I was wondering whether there are more appropriate techniques knowing that my input will be strictly in 2D ?

I am open to try another framework/technique providing it's not overly complex.

Here are source images :

enter image description here

Result of the 7 first images : (for some reasons, the Stitcher in OpenCV did not accept 10 at once ...)

enter image description here

Result of the last 3 images :

enter image description here

aybe
  • 15,516
  • 9
  • 57
  • 105
  • The question is much clear now. – karlphillip Jul 06 '12 at 02:16
  • For the Mario Bros example at least it would be difficult to do image stitching, as there is an aliasing problem (the pattern is repetitive). I suppose some sort of relatively high frequency feature tracking is necessary to register consecutive images in the video (said in other words, you cannot use images distant to each other otherwise you would have a lot of ambiguities) – Antonio Apr 01 '15 at 20:38

0 Answers0