extract fig into surf(x,y,z)

There's a .fig that plots n z(x) curves, each of which corresponds to 1 y value (n=3 in the illustration). The y values are known data. I don't have those z(x) data besides this .fig. How to convert this fig into a surf(x,y,z) plot?

[DGM]

Do you have an actual .fig file, or do you just have this sketch?

Even if you have a .fig file with internal data, it may not contain all the data necessary to unambiguously reconstruct the 2D function it represents. Consider the example:

% some 2D function
x = linspace(0,2*pi,100); % uniformly-spaced xdata
y = [0 0.4 0.6 0.75 0.9 1.05 1.2].'; % arbitrarily-spaced ydata
Z = sin(x).*cos(y); % Z = f(x,y)

% this is the surf that we might want
subplot(1,2,1)
surf(x,y,Z)

% but this is the only object we have
% if this were the only thing in the .fig,
% then the ydata would be lost.
subplot(1,2,2)
plot(x,Z)
grid on
xlim([0 2*pi])

In this example, our simple line plot represents a 2D surface, but the object created by plot() does not actually contain the ydata. Unless there were a legend or something which specifies the y-position of each of these curves, we can't know what they were. There may be exceptions in cases where we know some information about the function being described, but I'm going to assume we don't.

In order to know if the question is answerable, we need to know what information is available.

[feynman feynman]

Thanks a lot. Let's assume that the y values are known besides the .fig (not a sketch) data.

[DGM]

I reduced part of the above example to a fig file with nothing other than the 2D line part:

% say we have a fig file with some 2D line plots
hfig = openfig('stuff.fig');

% assuming it contains a number of 2D line objects
hl = findobj(hfig,'type','line');

% let's say we knew what the ydata was
y = [0 0.4 0.6 0.75 0.9 1.05 1.2].';

% assuming the xdata,zdata length is consistent across all line objects
xlen = numel(hl(1).XData);
nlines = numel(hl);
X = zeros(nlines,xlen);
for k = 1:nlines
    ki = nlines - k + 1; % expect object order to be reversed
    X(k,:) = hl(ki).XData;
    Z(k,:) = hl(ki).YData;
end

% go ahead and expand y in the case that all the xdata vectors are not identical
Y = repmat(y,[1 xlen]);

% make a new surf plot
figure
surf(X,Y,Z)

That gives us the same surf plot that we had before. If the data series aren't the same length, you might need to jump through a few other hoops.

[feynman feynman]

Thank you so much it works.

It even works when hfig isn't lines but dots. But if hfig is just dots, why does matlab still identify them as lines and group them into several clusters?

[Walter Roberson]

If hfig is dots but the code succeeds in creating a surface, then the implication is that the dots were created as line() objects with 'linestyle', 'none' instead of being created as scatter() objects. Creating dots as line() objects instead of as scatter() objects is fairly common.

The code gathers all of the XData and YData of the line() objects and ignores what the 'linestyle' is set to for the line() objects.

Answer 1

[Star Strider]

One way is tto enlarge the y-values to a matrix using repmat and then offset them by adding a different constant to each column (in this example).

x = linspace(0, 1).';
z = sin(2*pi*x*(1:3));
y = ones(size(z,1),1);
xm = repmat(x, 1, size(z,2));
ym = repmat(y, 1, size(z,2));

figure
plot3(xm, ym, z)
grid
colormap(turbo)
title('‘plot3’ Using ‘Original’ Y-Values')

figure
surf(xm, ym, z)
grid on
colormap(turbo)
title('‘surf’ Using ‘Original’ Y-Values')

ym = ym + (0:2);                                    % Create ‘Offset’ ‘Y’-Values

figure
surf(xm, ym, z)
grid on
colormap(turbo)
title('‘surf’ Using ‘Offset’ Y-Values')

.

[feynman feynman]

both answers are great. I don't know which to accept.

[Image Analyst]

@DGM posted his as a comment, not an Answer, so you can't accept his or even vote for it. Only @Star Strider posted an actual answer so unless DGM changes/moves his to an Answer you can only accept Star's.