Local smoothing of peaks for a matrix

Good morning,

I have a problem concerning the data of a matrix. Each data represents the value of curvature of a plate through its surface. However, in some regions of the surface (and then, in some group of values of the matrix) there are some strange values that I need to be smoothed. I can't filter out those values because I need all the values of the matrix, so it's just necessary to smooth those strange values. I tried to use the function "smoothdata", but the problem is that it smooths all the data of the matrix, while I need the correct values to remain not modified. I need some local smoothing/filtering.

You can find attached an image which show the plot of these curvature values through the surface.

Here is the code of this plot, KxxFE is the matrix of interest.

figure('color',[1 1 1])

surf(KxxFE);

colorbar;

colormap('hot');

Any help about this? Thank you so much

Answer 1

[Star Strider]

If you have R2018b or later, see if the rmoutliers function will do what you want.

[Luca Sulpizio]

Thank you very much for your answer.

I tried this command, but unfortunately it eliminates rows that contain outliers, but I need the matrix to remain the same dimension.. is there a way to make the outliers not to be deleted neither imposed equal to zero, but just similar to the "normal" values of the elements nearby to them in the matrix?

[Star Strider]

My pleasure.

The only other option I can recommend is the filloutliers function. There may be similar functions you can experiment with as well.

[Luca Sulpizio]

This function is way better, and it manages to solve the problem about dimensions, even if I need a more accurate smoothing, but I will eventually try with similar functions. Thank you so much!

[Star Strider]

As always, my pleasure!