Single affine parameter vector for whole image

asked 2017-06-04 02:28:23 -0600

UsmanArif gravatar image

I have an image of size 240x320. I want to calculate single affine parameter vector (6 parameter vector) for whole image using pyramids based Lukas and Kanade. The routine availbale in opencv calculates the dense model parameters, i.e. the parameters for all pixels separately, But i want one affine model for whole image. I have written my own routine but it slower, is there a builtin method in opencv..?

edit retag flag offensive close merge delete

Comments

btw, calcOpticalFlowFarneback does dense optical flow per pixel, the pyrLK version does not.

then, there is camera motion, and motion from (various) objects in the scene, all going into a possibly opposite direction. what's your plan here ? what exactly are you looking for ?

berak gravatar imageberak ( 2017-06-04 02:43:11 -0600 )edit

Actually I have images of a planar surface like floor and I want to calculate afine transform parameters between two floor images without using any feature points. Because in some situations we don't have feature points available. If there is another method for this problem, please tell. I don't need dense flow, I need one constant flow value(affine) for a single image.

UsmanArif gravatar imageUsmanArif ( 2017-06-04 09:02:48 -0600 )edit

have you tried phase correlation ?

berak gravatar imageberak ( 2017-06-04 09:07:36 -0600 )edit

No, in my case motion is large. Is that effective ?

UsmanArif gravatar imageUsmanArif ( 2017-06-04 09:48:43 -0600 )edit