The 2D PIV velocity field looks good from my data using the first tutorial notebook. Hooray!
But what I really want is an Eulerian field of the Lagrangian tendency: isolating the development of convection cells, wave modulations, etc. That is, I want to subtract from image2 an image composed of the image1's pixels, shifted according to the PIV-derived velocity (or really, displacement) field.
I can see ways how I could proceed, using the output coarse V field to shift and write sliced-out patches of values from image1 into a full-resolution dummy array. There would be gaps to fill where the flow is divergent, and overlaps to resolve somehow where it is convergent.
If I used buffer zones or "overlap" then I could avoids gaps and only have to resolve overlaps. As a programmer, I could do something from scratch.
winsize = 32 # pixels, interrogation window size in frame A
searchsize = 38 # pixels, search in image B
overlap = 12 # pixels, 50% overlap
But am I reinventing the mousetrap here? Does someone already have an elegant code? Or the right words to use to improve my google-fu about finding more?