3

I have 2 video files of a scene from two different angles and I want to reconstruct a 3D estimation of the scene. It is similar to what 3D sensors do (e.g. Kinect, PrimeSense). I am looking for a library or even a well-established machine vision algorithm so that I can implement an API myself. Thanks.

MostafaMV
  • 2,181
  • 3
  • 16
  • 22
  • Take a look at http://stackoverflow.com/questions/6570188/3d-model-construction-using-multiple-images-from-multiple-points-kinect?rq=1 – old-ufo Nov 10 '13 at 13:41

1 Answers1

2

You can use only one video file:

You can use Visual SFM (based on bundler and cmvs) to create a dense 3D reconstruction:

http://ccwu.me/vsfm/

You can also use PMVS in conjunction with Bundler to produce a dense point cloud: http://www.di.ens.fr/pmvs/ http://www.cs.cornell.edu/~snavely/bundler/

And there's also the SFM+CVMS distribution for windows (again, producing a dense point cloud): https://code.google.com/p/osm-bundler/downloads/detail?name=osm-bundler-pmvs2-cmvs-full-32-64.zip

GilLevi
  • 2,117
  • 5
  • 22
  • 38