Does not give satisfactory result with own data

[mcekwonu]

I used your code with my experimental data but it did not give any satisfactory results, unlike using your generated synthetic schlieren images.
I have attached samples experimental images. I am still investigating why it failed. Could you help?

[kasper64]

Hey Michael,
Thanks for reporting this issue.
The images you provided are in colour, and apparently the code does not handle this. I will add the conversion to monochrome. If you are in a big hurry, you can convert the images to monochrome manually.

With kind regards,
Kasper

[mcekwonu]

Ye
Yes, they are colored but I have already edited the code to read and load the images in grayscale using opencv "as cv2.imread(filename, cv2.IMREAD_GRAYSCALE)", so the image is two dimensional ( 1024x512).
My main issue is why I am not getting the density field. This is the result I got.

[kasper64]

Ah I see.
I got the same result as your attached image.
I'm not very knowledgeable about the FCD process, but I think a issue could be that your reference pattern seems to be shifted so much that it makes small changes in pattern phase irrelevant. When it creates the resulting picture, the data from FCD is normalized, which washes out the small changes in phase.
I tried to align your pictures and apply the FCD process. Now I get the result attached.

I'm interested what output picture you would expect.

[mcekwonu]

I see. Actually, the images are experimental images of BOS that I conducted for the spray jet from the nozzle, and the reference image was taken when there was no spray and a distorted image when the spray jet was turned on, so I expected to see the flow fields of the spray jet. It means that I should remove the normalization of the height_field.
I cloned your code and then modified it using OOP for ease of usage, but I am most concerned about the output, which is not satisfactory compared to using PIVlab (DIC method).
I will try it again when I get home, but the result you display looks a bit closer but is not yet revealing the flow field.

Thank you

[kasper64]

To check whether the problem lies in the FCD method or my implementation, you could give the original Matlab FCD algorithm a try with your data. See readme.md for a link. Unfortunately I lost access to Matlab so I can't easily try it.

[mcekwonu]

I am studying Wildeman algorithm and checking your implementation in python. I will try the MATLAB code. I kind surprised how it worked with your generated image but failed with mine.

[kasper64]

Hey hey,
Did you find out more? I'm curious.

[mcekwonu]

I am rewriting the codes using swilderman Matlab codes to test with my experimental data. I will update and share with you as soon as I finish it. If you don't mind, we could collaborate to make it a package and publish on pypi. Michael C. Ekwonu, PhD

…

On Sat, Feb 24, 2024, 5:33 PM kaspervn ***@***.***> wrote: Hey hey, Did you find out more? I'm curious. — Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEINKXFVQJNQVMSG6ZT7J7TYVGQUBAVCNFSM6AAAAABC3KBW4KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNRSGI4TQNBVGA> . You are receiving this because you authored the thread.Message ID: ***@***.***>