Filter recurring oscillation in 2D Data

Hi,

I am creating a 2D surface plot where I can see a diagonal pattern in my data. I see recurring oscillations which I would like to filter out from my original signal.

Andy advice on how this could be done without disturbing the original signal too much? I tried smooth, median, etc. without any success.

Thanks.

[William Rose]

@Konvictus177, I do not see a diagonal oscillation in the image. I assume the plot is z(x,y). The z values appear to be identical at all y values, and z oscillates in the x direction. Therefor your image seems to be inconsistent with the question you have posed.

Show the code you tried, which did not produce the results you expected.

[DGM]

There is a very faint diagonal banding, though depending on how this plot/image were generated, it might just be aliasing due to display interpolation and the dominant features being slightly angled WRT the figure.

Answer 1

[William Rose]

@DGM, @Konvictus177, I cannot see diagonal banding. The arrows added to th image point to a vertical line, not a diagnonal feature.

If I had unwanted parallel diagonal fluctuations across the entire image, I would do a 2D FFT, then zero out the unwanted frequencies on the appropriate diagonal in the frequency domain, then do the inverse 2D FFT.

If I had a single unwanted parallel diagonal fluctuation going across the image, I would try to fix it in the spatial domain. I would set up a for loop to proceed along the cenerline of the diagonal, replacing each point with the median of itself and the N points in a perpendicular direction on either side. Then I woudl repeat for the parallel set of points on either side of the centerline, etc. There are variaiton so on this approach that you could try, such as replacing 3 or more perpendicular-to-the-diagonal points at a time with the median of those 3 or more points.