How to Output Accurate PDF from Matlab figure

Hi everyone:

For some reason, I need to plot the mask of a microfluid device in Matlab, which consists of many posts separated by certain distance The plotting process is easy but when I tried to 'print' it into PDF file, the image was distorted (circles are not round, and they have obvious defect on them), and the spacing was not right(shortened).

I've tried function 'rectangle', code like:

function drawFilledCircle (center_x, center_y, radius)
%%Draw a filled circle 
xc = center_x; yc = center_y;
r = radius;
x = r*sin(-pi:0.01*pi:pi) + xc;
y = r*cos(-pi:0.01*pi:pi) + yc;
fill(x, y, [0,0,0]);

and even a function called 'filledCircle' which I found on fileexchange.

But I always got the wrong thing. I don't know what I can do to fix it.

Please help.

P.S. I attached the original matlab figure and the printed PDF for references.

P.S.2 I've also tried to output them into other formats like EPS, but even EPS doesn't look right.

[lvn]

I tried zooming in on a small part of your figure (it seems to consist of hexagons made up from circles) and exporting it to pdf.

print -dpdf test1.pdf

I attach the result which looks to be ok.

[Frank]

If I zoom in I have no problem get the right feature. But If you want to print the whole figure and you zoom in after you print it, it's always wrong.

[lvn]

Could you post the complete code that would reproduce test.fig? I would like to see if I manage to export it correctly.

[Frank]

No problem. Here you go, thanks. (Please let me know how to do it if you manage to do so)

%%Hexagon in Hexagons

%%if we want to do size of 7 (hexagon in a hexagon array)
x_num = 30;
y_num = 30;
diameter = 100*7 + 100*6;
distance = diameter;
margin = 100;
% Define the long distance in the hexagonal lattice
long_distance = sqrt(3)*(distance + diameter) - diameter;

xlimit = 2*margin + diameter*x_num + (x_num-1)*(distance - diameter)/2;
display(xlimit);

ylimit = 2*margin + diameter*y_num + (y_num-1)*(sqrt(3)*(diameter + distance)/2 - distance);
display(ylimit);

figure;set(gca,'position',[0 0 1 1]);
set(gca,'DataAspectRatio',[1,1,1]);
% set(gca,'XColor',[1,1,1],'YColor',[1,1,1]);
set(gca,'XTick',[],'YTick',[]);
set(gca,'xlim',[0,xlimit],'ylim',[0,ylimit]);
box on;
% for x = 0:x_num-1
%     for y = 0:y_num-1
%         rectangle('position',[margin+(x*(diameter+distance)),margin+(y*(diameter+distance)),diameter,diameter],'curvature',[1,1],'FaceColor',[0,0,0]);
%     end
% end

%Try to plot the posts by even and odd 

if mod(x_num,2) == 0
    odd_num_x = x_num/2;
    even_num_x = x_num/2;
else
    odd_num_x = (x_num - 1)/2 + 1;
    even_num_x = (x_num - 1)/2;
end

if mod(y_num,2) == 0
    odd_num_y = y_num/2;
    even_num_y = y_num/2;
else
    odd_num_y = (y_num - 1)/2 + 1;
    even_num_y = (y_num - 1)/2;
end

for x = 0:odd_num_x-1
    for y =0:odd_num_y-1
%         rectangle('position',[margin+ x*(diameter + distance) , margin+(y*(diameter+long_distance)),diameter,diameter],'curvature',[1,1],'FaceColor',[0,0,0]);
        draw_hexagon (margin+ x*(diameter + distance) + 250, margin + (y*(diameter+long_distance)),7,100,100);

      end
  end

for x = 0:even_num_x-1
    for y = 0:even_num_y-1
%         rectangle('position',[margin+ (distance + diameter)/2 + x*(diameter + distance) , margin + sqrt(3)*(diameter + distance)/2 + (y*(diameter+long_distance)),diameter,diameter],'curvature',[1,1],'FaceColor',[0,0,0]);
        draw_hexagon (margin+ (distance + diameter)/2 + x*(diameter + distance) + 250, margin + sqrt(3)*(diameter + distance)/2 + (y*(diameter+long_distance)),7,100,100);

      end
  end

The function draw_hexagon I use above is:

%%Function to print hexagonal shaped island in a blank area
  function draw_hexagon (position_x, position_y, size, diameter, distance)

    % Size must be odd.
    % Size is the number of posts on the longest diagonal of the hexagon.
    if mode(size,2) == 0
        error('Please input an odd number for the size parameter!');
    end

    side_num = (size - 1)/2 + 1;

    % The strategy is to plot the lower half of hexagon, and then plot the top
    % half of the hexagon.

    % define a distance parameter that can switch between top and bottom
    up_down = (side_num - 1)*sqrt(3)*(distance + diameter);

    while side_num <= size
       for index = 0:side_num - 1
           rectangle('position',[position_x + index*(diameter + distance), position_y, diameter, diameter],'curvature',[1,1],'FaceColor',[0,0,0]);
           rectangle('position',[position_x + index*(diameter + distance), position_y + up_down, diameter, diameter],'curvature',[1,1],'FaceColor',[0,0,0]);
       end
        side_num = side_num + 1;
        up_down = up_down - sqrt(3)*(diameter + distance);
        position_x = position_x - (distance + diameter)/2;
        position_y = position_y + sqrt(3)*(distance + diameter)/2;
    end

    % positions are the lower left corner of the hexagon

    end

Sorry It's not well commented. I was not intended to be distributed. Please let me know if you have any problem reading it.

Answer 1

[lvn]

The problem with the spacing is in fact a problem of the line around the circle. Because you work on such small scales, it determines the size of the circle! This solves it I think. In your code change the rectangle functions to

rectangle('position',[position_x + index*(diameter + distance), position_y, diameter, diameter],'curvature',[1,1],'FaceColor',[0,0,0],'LineStyle','none');

and then Jan Simon's export print(gcf, 'test.eps', '-dpsc2', '-r1200', '-noui').

After conversion to pdf, the result looks ok to me (see attached).

[Frank]

Thanks a lot Ivn! It works for me now. I'll let you know if I have more problems using LineStyle.

One more question: Why do we need to output into eps and convert into pdf? I just checked out my 'direct output' pdf, it also looks fine.

[lvn]

Glad it works! No particular reason for eps I guess (I never use pdf).

[Frank]

Add one thing to whom may concern:

In my case, I want to output nearly 10^5 posts on one figure, it's important to set the resolution to at least '-r1600' to obtain undistorted image.

Answer 2

[Chad Greene]

I highly recommend switching to export_fig instead of using print. Here's export_fig: http://www.mathworks.com/matlabcentral/fileexchange/23629-exportfig

Oliver Woodford, the author of export_fig has a quite helpful and succinct site with plenty of picture examples and troubleshooting tips here: https://sites.google.com/site/oliverwoodford/software/export_fig

Does the attached pdf look correct?

[Frank]

Hi,

I've also tried export_fig but still get the wrong thing. Maybe you can teach me how to set up the parameters in export_fig.

BTW, the PDF i attached looks wrong (and I don't know how to get it right). You can zoom in and compare to .fig, then you will know what a right one look like.

[Chad Greene]

I'm still not exactly sure what the problem is; perhaps it's an issue with your pdf renderer? Does this png look correct?

[Frank]

FYI, this is the PDF I output by export_fig. And it looks no better than what I attached above. Please check it out.

[Frank]

Let me put it in a clear way. In my matlab figure file attached above, the diameter of the post is 100um, and the spacing between two posts are 100um as well. Did any of the output file (PDF,PNG) has perfect round posts and right spacing?

Answer 3

[Chad Greene]

Oh, interesting. I see the problem now, but is this a pdf issue, or an issue with the filled circle function? Perhaps simply creating a scatter plot and setting the size of the markers would fix it? http://www.mathworks.com/matlabcentral/answers/101738

[Frank]

Good idea, it's really smart. I'll try it out, but at this point I don't know whether this will solve the problem.

My intuitive guess is it has something to do with the rendering. I tried '-painters', '-zbuffer', and '-opengl', but none of them improve the quality. Also I've tried to play with the resolution '-r', but the distortion is still there even if I increase the resolution to very high.

[Frank]

I tried to output by scatter. The result is better since the circles are round. But the problem is the marker size on figure doesn't match the size on pdf. It seems that matlab automatically scale it when you output figure into pdf. Here are the figure and pdf I got.

Answer 4

[Chad Greene]

Frank,

Perhaps this will help? http://www.mathworks.com/matlabcentral/fileexchange/45952-circle-plotter

Follow the circles command with axis equal to ensure the circles don't get squashed.

[Frank]

Thanks for your code. I tried it out, and the result is weird because: not only the spacing is still not right, but also the circles are defected.

I my original script, I already use

set(gca,'DataAspectRatio',[1,1,1])

to set the axis scaling factor to 1 (which works the same as

axis equal)

BTW, I tried the scatter plot, the dots look better than plotting circles, but the size control (SizeData) method doesn't count for the PDF.

I think we may keep thinking about scatter plot, or consider if something is wrong with the rendering issue.

[Frank]

Hi Chad, the results of scatter plot is very promising except i don't know how to control the size of the marker.

According to the link you send to me,I use:

s = 100; %Marker width in units of X
currentunits = get(gca,'Units');
set(gca, 'Units', 'Points');
axpos = get(gca,'Position');
set(gca, 'Units', currentunits);
markerWidth = s/diff(xlim)*axpos(3); % Calculate Marker width in points

to get the size of the marker, and then use:

scatter(X,Y,makerWidth^2)

to plot. The circles are not distorted on the output pdf, however, the size is apparently wrong. Do you have any idea on size control?

Answer 5

[Jan]

This creates a perfectly looking EPS and after a conversion the resulting PDF is fine also:

print(gcf, 'test.eps', '-dpsc2', '-r1200', '-noui')

Increasing the resolution to 1200 makes the difference.

[Frank]

Hi Jan,

Thanks for your effort. The circles look perfectly round. But the other issue still exist : the spacing between posts are definitely not right. (if you look at my code, then you can see spacing = diameter)

I don't know if it has something to do with rendering. But I'm happy to discuss with you anytime if you are willing to do so.