General idea: calculate optical flow vector fields (LK coarse to fine for stability?) -> normalize each field back into gray images -> use histogram thresholding to retrieve foreground of moving cilia -> obtain dynamic masks for each video with hopefully minimal false positives (yielding highly textural samples for appearance head input)
Scripts will be in folder 'preprocessing' & progress will be tracked via this ticket!
General idea: calculate optical flow vector fields (LK coarse to fine for stability?) -> normalize each field back into gray images -> use histogram thresholding to retrieve foreground of moving cilia -> obtain dynamic masks for each video with hopefully minimal false positives (yielding highly textural samples for appearance head input)
Scripts will be in folder 'preprocessing' & progress will be tracked via this ticket!