CODEWITHSUNDEEP
c++opencv
Alex
Alex

Reputation: 51

Motion estimation for video sequences

For my project I have to track the movement of some objects in a video sequence, for motion estimation using a technique that is based on an algorithm of block-matching between two consecutive frames, at this point I would like to use the information on motion estimation obtained at the previous frame to produce that of the next frame, that is, to use the information on the previous estimate for the current estimate without having to process strength for two consecutive frames of the sequence. Can someone suggest me some technique??

thanks

Upvotes: 0

Views: 1333

Answers (2)

rotating_image
rotating_image

Reputation: 3076

Use optical flow technique...use goodfeaturestotrack to find some points of interest and keep tracking them in each frame...the displacement of these points in each frame compared to the previous frame will give you some idea about the nature of motion between 2 frames..

Upvotes: 1

akrami
akrami

Reputation: 309

if you want to track something in a video, use mean-shift and cam-shift. opencv provides these two algorithms. you can use cvMeanshift or cvcamshift functions for tracking. the algorithm is based on center of mass calculation. mean shift finds the center of mass of the given object that should be tracked then finds it in the next frame by an optimization algorithm. in cam-shift approach if the dimensions of the object change, it can already track it.
you can find more informations in opencv website about mean-shift and cam-shift

Upvotes: 0

Related Questions