How to remove outliers in a 3D surface

I have a 3D surface (see below) construced using surf command based on excel data:

T = readtable('https://in.mathworks.com/matlabcentral/answers/uploaded_files/791474/excel-surface.xlsx')

x = T.(1) ;

y = T.(2) ;

z = T.(3) ;

xi = linspace(min(x),max(x));

yi = linspace(min(y),max(y)) ;

[X,Y] = meshgrid(xi,yi) ;

Z = griddata(x,y,z,X,Y) ;

surf(X,Y,Z)

shading interp

colorbar

The resulting surface is as follows:

However, I know that the surface should be smooth without these "peaks" that appear randomly from above or below.

Is there a way to remove these outliers and interpolate to construct a smooth surface? (NOTE, I AM NEW TO MATLAB).

Answer 1

[Matt J]

You could try rmoutliers().

[HAMID]

can this be done for a surface? (i am completely new to matlab)

[Matt J]

Yes, you can do 2D outlier removal.

[HAMID]

I am trying to use the method, but I am confused. I am new to this software.

[Matt J]

I would have to see what you tried in order to be able to guess what's not working.

[HAMID]

I tried fill outlier instead of rmoutliers(), but I am obtaining this:

https://www.mathworks.com/matlabcentral/answers/1580479-how-to-remove-outliers-in-a-3d-surface#comment_1819454

Answer 2

[Chris]

I would recommend trying the "clean outlier data" task in the live editor, after sorting the data:

T = sortrows(T,1); % Sort by the first column ('x')

Open a New Live Script, and select the task.

Select the table and input variable z.

I can get decent results with spline interpolation, a centered moving median, and a low threshold. But I'm not sure how, computationally, you would create a perfectly smooth surface that retains the contours you would want to see, as there are quite a few clumped outliers.

Perhaps repeated sorting by x and y, with outlier cleaning in between, could work.

[HAMID]

clear; clc;

% Read table:

T = readtable('https://in.mathworks.com/matlabcentral/answers/uploaded_files/791474/excel-surface.xlsx')

% Fill outliers:

T = sortrows(T,1);

T.Z = filloutliers(T.Z, "spline", "movmedian", 4);

T = sortrows(T,2);

T.Z = filloutliers(T.Z, "spline", "movmedian", 4);

% Plot surface:

x = T.(1) ;

y = T.(2) ;

z = T.(3) ;

xi = linspace(min(x),max(x));

yi = linspace(min(y),max(y)) ;

[X,Y] = meshgrid(xi,yi) ;

Z = griddata(x,y,z,X,Y) ;

surf(X,Y,Z)

shading interp

colorbar

I tried sorting multiple times but it is not working. I now have a new weird outlier.

Before:

After:

[Chris]

clear; clc;
% Read table:
T = readtable('https://in.mathworks.com/matlabcentral/answers/uploaded_files/791474/excel-surface.xlsx');
% Fill outliers:
for idx = 1:2
    for jdx = 8:-2:2
        T = sortrows(T,1);
        T.Z = filloutliers(T.Z, "spline", "movmedian", jdx);
        T = sortrows(T,2);
        T.Z = filloutliers(T.Z, "spline", "movmedian", jdx);
    end
end

% Plot surface:
x = T.(1) ;
y = T.(2) ;
z = T.(3) ;
xi = linspace(min(x),max(x));
yi = linspace(min(y),max(y)) ;
[X,Y] = meshgrid(xi,yi) ;
Z = griddata(x,y,z,X,Y) ;
surf(X,Y,Z)
% shading interp
colorbar

Getting pretty close.

With default shading, you'll see there are still a few bumps. But you can also see interesting features. Repeating the process, and changing the window size, continues to smooth the surface.