Conditional Removal of points from triangular matrix shown by Contour in Matlab

Question

I have big triangular matrix.

enter image description here

which has relevant data marked by different colours. I want to remove conditionally the points marked by greenish-yellowish contours:

Remove greedily on the area of non-linear objects
Remove lazily on the area of linear objects with probability 50%

My attempt

I run based on Shai's comment

% remove linear things on nonlinear area lazily: matrix(97:103, 1:98)
% remove linear things greedily elsewhere
for row=0:97
    for column=0:111
        % Lazy removal
        if and(row > 97, row < 104)
            if and(column > 0, column < 98)
                if randn > 0
                    matrix( matrix < 9 ) = 0;
                end
            end
        end

        % Greedy removal
        if or(column < 97, column > 104)
            % Remove all points in these regions because no linear objects here
            matrix(:, 1:97) = 0;
            matrix(:, 104:111) = 0;
        end
    end
end

I get

enter image description here

which is lot better than the unconditional removal

enter image description here

but still the conditional part of lazy removal can be improved. I think you cannot use here Shai's shorter version, and nested loops must be used because you have conditional removal.

You cannot use contour lines, like contour(matrix, clines) because the non-linear objects cover the linear objects too. So you need conditional removal by selecting specific area of the figure for greedy removal and lazy removal. Daniel R's command, contour(...,'ShowText','on'), does not seem to help us here, and we cannot simply remove by value. I think the following figure shows the zero points, probably singularities , because there should be 111 singularities in the figure. Does the following figure show singularities or only zero values of the data?

enter image description here

How can you apply a specific removal rule to the area of linear objects?

How can you remove conditionally the points that are marked by greenish-yellowish color in the triangular matrix in Matlab?

What kind of plot is this? Contour plot? Then use the input parameter (v) to specify which levels are shown. — Daniel, Nov 17 '13 at 13:55
I developed the idea of conditional removal in this thread further: http://stackoverflow.com/questions/20168756/to-apply-a-function-through-a-mask-on-an-image It deals with the random logical matrix too better. — Léo Léopold Hertz 준영, Nov 26 '13 at 17:45
Please, close this question because the same issue is targeted in the other post. Also, I cannot maintain the original data set anymore. — Léo Léopold Hertz 준영, Nov 26 '13 at 17:47

Robert Seifert · Accepted Answer · 2013-11-18T15:17:28.277

2

Your data is complex, Knowing that, and knowing what the contour function is doing with complex numbers - just plotting the real part - it is easy to filter your data.

Just set every element of your matrix to NaN, your threshold you can find out by just plotting a colorbar with your original plot and guess the values. Or have a look at the real part of your data matrix, etc.

Adjust the threshold as desired

Be aware that the colorbars should be equal for comparison

load('tri4_good.mat');
%original plot
figure(1)
[~,h] = contour(samii);
colorbar
caxis([-2,1.5].*10^7)

%get data
X = get(h,'Xdata');
Y = get(h,'Ydata');
Z = real(samii);
%this plot equals the original one
figure(2)
contour(X,Y,Z)
colorbar
caxis([-2,1.5].*10^7)

%therefore the Z-Data is equal to the real part of your data
%X and Y are the indices of your data matrix
%set the threshold as desired
thresh = 0.1*10^7;

idx = find( abs(Z) < thresh);  %or what condition you like to use.
Z(idx) = NaN;

%filtered plot
figure(3)
contour(X,Y,Z)
colorbar
caxis([-2,1.5].*10^7)

gives you:

enter image description here

regarding all your comments, these lines should be the right way to go:

Z = abs(samii);
idx = find(Z == 0);
Z(idx) = NaN;

%filtered plot
figure(1)
contour(Z);
colorbar
caxis([-2,1.5].*10^7)

returns:

enter image description here

If this doesn't fit your needs, you have to completely overthink everything ;)

edited Nov 18 '13 at 15:17

answered Nov 18 '13 at 14:33

Robert Seifert

25,078
11
68
113

What is the contour function doing with complex numbers? I need complex numbers in my application. So probably contour is not then best way to filter the nonlinear objects out from my data. What other methods there exist? – Léo Léopold Hertz 준영 Nov 18 '13 at 14:47
what do you consider as non-linear? I mean instead of Z = real(samii) you can set Z = imag(samii). And it will plot your original data just with the imaginary part. – Robert Seifert Nov 18 '13 at 14:49
I have an intuition that the contour function is doing abs of the whole data before plotting contours. – Léo Léopold Hertz 준영 Nov 18 '13 at 14:50
Nonlinear is nonlinear: http://en.wikipedia.org/wiki/Nonlinear_system Ok here in my data, I have already filtered most of it away. It it was heavily nonlinear at the region which I have described here nonlinear. – Léo Léopold Hertz 준영 Nov 18 '13 at 14:52
It takes the real value, not the absolute value. You can check it by just comparing `real(samii)` with `abs(samii)` – Robert Seifert Nov 18 '13 at 14:55
well, how can your matrix tell me what is non-linear, there are basically just values inside. – Robert Seifert Nov 18 '13 at 14:57
Then, I have to think again how to filter data better. I think the contour show well the regions where the linear objects occur. Still, I think this is enough in my case. **Is there any other way to visualise complex data that show similar situation?** – Léo Léopold Hertz 준영 Nov 18 '13 at 14:57
I cannot answer your question about nonlinearity here because of limited space and time. – Léo Léopold Hertz 준영 Nov 18 '13 at 14:58
Assuming you get the complex values from derivates, than `Z = abs(samii)` and filter these values with threshold or setting all zero elements to `NaN` would be the right way to go. – Robert Seifert Nov 18 '13 at 15:06
Thank you for your answer! You have very good points in your answer! – Léo Léopold Hertz 준영 Nov 19 '13 at 07:17
See this thread http://stackoverflow.com/questions/20168756/to-apply-a-function-through-a-mask-on-an-image I developed the idea of conditional_removal and randn_logical_operator_matrix. – Léo Léopold Hertz 준영 Nov 26 '13 at 17:44

score 1 · Answer 2 · answered Nov 17 '13 at 14:03

1

A simple way, you can just change the value of a matrix element, when it is not equal than VALUE change it 0. Ask a green/yellow color number VALUE in the color scheme, say it is VALUE=125.

for i=1:size(matrix,1)
   for j=1:size(matrix,2)
     if(matrix(i,j)==125)
        matrix(i,j) = 0;
   end;
end;

answered Nov 17 '13 at 14:03

Plo_Koon

2,953
3
34
41

2

Ahh!!!! why nested loop?? `matrix( matrix == 125 ) = 0;`. Moreover, it is [best not to use `i` and `j` as variable names in Matlab](http://stackoverflow.com/questions/14790740/using-i-and-j-as-variables-in-matlab). – Shai Nov 17 '13 at 14:08
Very good points! Do you have any idea what is the color code of the greenish data? Or how to find the value of the level of data? Is there any possibility to select a data point and this way see the value? – Léo Léopold Hertz 준영 Nov 17 '13 at 14:38
1

contour(...,'ShowText','on') – Daniel Nov 17 '13 at 16:17

Conditional Removal of points from triangular matrix shown by Contour in Matlab

My attempt

2 Answers2