narain, 1 month ago @demofox It's basically a edge detection filter / finite difference gradient (convolve with [-1, 0, 1]) except applied in time rather than space. Which suggests to me that convolving with [-1, 2, -1] (i.e. do sharpening instead of edge detection) may give more naturalistic results...
@demofox It's basically a edge detection filter / finite difference gradient (convolve with [-1, 0, 1]) except applied in time rather than space.
Which suggests to me that convolving with [-1, 2, -1] (i.e. do sharpening instead of edge detection) may give more naturalistic results...