Interpolation by increasing window overlap

[Jonathan S.]

Dear PiVlab users,
I want to increase the resolution of my PIV as much as possible (potentially 1 pixel). The reason is that I am trying to feed concentration and velocity field at every point (x,y) into a PDE-learning algorithm. If my PIV field has a bad resolution, then i can feed in the concentration field only the same bad resolution. 
I have tried interpolating the velocity field, but I was thinking that the best form of interpolation is probably using a step size of 1 for the final PIV pass. With normal interpolation the gradient of the velocity field (at the points that are interpolated) is not very meaningful - it is completely dependent on the interpolation technique. A small step size should do a better job.
Would you agree on that point?
In the PIV implementation the step size of the final step seems fixed however. Is there an easy way to change it? I have tried doing it manually in the function PIV_fftmulti (changing the value of step in multipass ==4), but it crashes, probably because other dependencies can not deal with a step size not being half the window size. Without reading deeply into the code, I was wondering if there is a quicker workaround and if you agree that this is a good idea.
Thanks a lot for your help!
Best wishes,

Jonathan

[William]

Dear Jonathan, as far as I remember, the book "PIV - a practical guide" shows an image of the vorticity of a vortex pair for different overlap percentages. And the result was that more overlap gives worse gradient. Please check the book.

There is no easy way to change the overlap percentage, as my window deformation assumes 50% overlap. But why not use the wOFV algorithm which natively gives 1-pixel resolution?

[Jonathan S.]

Dear William,

thank you so much for your help. This helped me so much! The section in the book was super useful. I am surprised how well the wOFV algorithm works for my problem with doing PIV on filaments.

Best wishes,

Jonathan