Smath does not take submatrix reversed cols order, to mirror. Alternately, use mirror(M).
The Venturi image as captured by a digital RGB camera is symmetric, though it may not look so after the processing. Engineer will isolate the trace most relevant of the continuous path of the fluid. The search is done manually, much less difficult than one might think.
Select/collect a path
=>
<=
Selected contour for user to extract and fit NON-linear.
1rst run Thiele on 23 discrete samples. It suggests a 2nd Thiele run, RemToDo.
This one, can't be expanded symbolic "confrac". Assuming a 2nd Thiele will succeeds, a linterp/cinterp might well conclude the project.
The silent difficulty, explain: The grayscale is in integer values, that makes non convenient to fit vs collected data from phenomenon. Nothing can cure or help otherwise, except from successive Thiele refit [jmG]. However, the "silent helper" is the easy contour, not noisy from epsilon ε = 1.
A comfortable ExpDecay2 fit
Module xd(L,H,N) is a discretiser for use in-situ [user spec's]: range [L,H], density [N].
Model function
Range & populate the discrete fit
Most interesting: the ExpDecay2 exposes point [8] misaligned. It's this point that makes Thiele like fail around this point.
jmGiraud:2016_02_14
Highlight by negative pixelation
experiment around A = -1.75 xo = 127
Plug in your system READ_IMAGE
Negative pixelation is misleading ... observe