Open DICeuser22 opened 3 years ago
feature matching: matches features throughout the global image from one frame to the next, then does a least squares fit of the motion of the features and assigns an initial guess of the subset displacement based on this least squares fit. feature matching with thresh: similar to the feature matching initializer above, but does applies a threshold to the image first to make it a binary image prior to searching for features. field values: use the displacement value from the last frame as the initial guess with no projection for the current frame (this is the default) image registration: this is useful in cases where there is large rigid body motions between frames. The initial guess for this method is determined by doing a global image registration between frames. neighbor values: the first subset uses its previous value for displacement as the initial guess, but every other subset uses a neighbor's displacement value in a sequential fashion starting with subset 0.
Thanks!
Could you explain the different initialization methods? I don't know the difference between them