1

I would like to ask if it is possible to deal inside openCV with the registration of an virtual object into a real image(real world object). After detecting the region of interest in a real captured image frame, I would like to substitute the pixels of the real image by the pixels of a virtual object which should appear as a real part of the new generated image.

plz he

Samarth Singhal
  • 11
  • 3

1 Answers1

2

For sure, (almost) everything is possible if you program it by yourself. However if you expect an on the shelf solution from OpenCV it doesn't exist... What you are talking about is called : pose-estimation

Depending on the context of your problem, it can be very difficult to do (depending on your computer sciences skills as well)

Instead of a very very long explanation, I think the best is to look at this :

You should try to look at what the field of virtual/augmented reality is, i think it could answer some of your questions... I don't have better answers as your question is very very wide.

Moreover a last tip would be to look at features detection and extraction as a lot of these techniques rely on a good detection of keypoints (to then replace a 2D-3D model into the scene)

Julien,

jmartel
  • 2,771
  • 4
  • 24
  • 27
  • We were trying to do it by SURF...we got the interest points...and if we could somehow map these ipoints to a predefined 3d model..it can work(maybe) – Samarth Singhal Jul 22 '11 at 08:27
  • Well... SURF is ok, even if for big images it is quite long and you will need an optimized version.. or more, a GPU version if you are lucky enough to use a GPU card. What do you expect more from my answer ? – jmartel Jul 22 '11 at 12:33