2020-04-25 08:16:28 -0600 | received badge | ● Famous Question (source) |
2019-07-26 04:28:15 -0600 | received badge | ● Good Question (source) |
2016-10-08 20:30:39 -0600 | received badge | ● Notable Question (source) |
2015-06-22 04:48:00 -0600 | received badge | ● Popular Question (source) |
2014-10-09 07:17:02 -0600 | received badge | ● Nice Question (source) |
2013-02-04 11:35:11 -0600 | commented answer | realtime video stabilization yes, this should work. since I am looking at frame by frame changes I can drastically reduce the search space. wow, so simple, yet so beautiful :). |
2013-02-04 07:26:54 -0600 | commented answer | realtime video stabilization response to edit: ah, ok, get it. though this only works for shifts. in reality there is also rotation and 3d effects, though maybe this is enough. thanks a lot! |
2013-02-04 07:24:43 -0600 | received badge | ● Supporter (source) |
2013-02-04 05:12:29 -0600 | received badge | ● Student (source) |
2013-02-04 04:34:21 -0600 | commented answer | realtime video stabilization Hi, thanks for the answer. I dont understand what you do in step 4. How is your voting set up? I never saw hough voting to match point sets. Also, how do you match? Do you extract descriptors? I was thinking about using ORB. is there a particular reason you do not use the matchers provided by opencv? 1ms sounds very nice! |
2013-02-03 05:10:46 -0600 | asked a question | realtime video stabilization Hi, We need a real-time method of video stabilization that does not hog the cpu too much. we also have a gpu in the machine and already do un-distortion and soon also background subtraction there. the camera is fixed, and many features are constantly visible, if that helps. do you have any hints on where best to start to solve this? I have been playing with the stabilization module, but found it to be orders of magnitude too slow. also, does it work online? Very thankful for any hints, and great work btw! without opencv we wouldnt be close to where we are now (soccercam.nl). Daniel |