Make figure without white border

I have a piece of code that plots a surface, and I feel it is ready for being included into the report. The only thing is that the figure has a quite large white border which substantially increases the size of it. I would like to have the unnecessary white space reduced to a minimum.

I did try

set(gca,'LooseInset',get(gca,'TightInset'))

But this cuts off my z-axis label. Is there an easy way to achieve this? With LaTeX, we can simply use the standalone documentclass with PGFplots/tikz, but I don't think there's a simple keyword or option to do this with matlab. I'm using R2016b.

Answer 1

[Chad Greene]

export_fig removes the border by default.

[Sim]

Hi, I have used both the following functions to save a figure as a PDF file:

print (figure_name, '-dpdf','-opengl','-r600', [path, plot_name{:}] ) 
export_fig([path, plot_name{:}], '-pdf', '-opengl','-r600');

Result:

Even though the matlab built-in function print does not crop the white margins in a PDF figure (right?), it mantains a very good resolution/quality of the saved PDF figure. The file exchange export_fig crops instead the white space in a figure automatically, but it reduces a bit the quality of the saved PDF figure, resulting in slightly blurry contours (after zooming in).

This is just my experience, but if you know a reason for the loss of quality in a PDF figure by using export_fig, please let me know!

All the best

Answer 2

[atilla Maia]

I have tried many suggestions. This one works for me:

f = figure;
F= getframe(f);
img = F.cdata;
imwrite(img, filename) ;

Answer 3

[Gyaneshwar dubacharla]

Use this command from MATLAB - exportgraphics(gca,fname.emf,'BackgroundColor','none')

Answer 4

[Hajra Afzal]

@Gyaneshwar dubacharla exportgraphics(gca,fname.emf,'BackgroundColor','none') worked for me. Thankyou for sharing. I faced quite a hassle before implementing this command.

Answer 5

[Jose Marques]

You can "imwrite" too.

Answer 6

[Thomas]

Maybe this way?

fig_1 = figure(1);
clf();
n = 512;
m = 2 * n + 1;
j = ones(1, m);
i = (-n : n) / n;
i = i .* i;
ii = i' * j;
img = real(sqrt(1 - (ii' + ii)));
imshow(img);
hold on;
lw = .01 * m;
x = 4 * lw + (m - 8 * lw) * [.25 .25 .50 .75 .25 .75 .25 .75 .75];
y = 4 * lw + (m - 8 * lw) * [1.0 .50 .00 .50 .50 1.0 1.0 .50 1.0];
plot(x, y, LineWidth = lw, Color=[1 0 0]);
title("ball with santa's house");
hold off;

new_bordersize = 2; % must not be greater than the original border
img = remove_border(frame2im(getframe(fig_1)), new_bordersize);
imwrite(img, "ball_with_santas_house.tiff");

%%REMOVE_BORDER
function img = remove_border(img, new_bordersize)
%REMOVE_BORDER
    % default value for new_bordersize is zero.
    if ~exist("new_bordersize", "var")
        new_bordersize = 0;
    end
    % get the border color
    col = img(1, 1);
    % get the image size
    [h, w, ~] = size(img);
    % search from left for the first column that not contains only the border-color
    x0 = 0;
    while x0 < w
        x0 = x0 + 1;
        if ~all(img(:, x0) == col)
            break;
        end
    end
    % search from rigth for the first column that not contains only the border-color
    x1 = w + 1;
    while 1 < x1
        x1 = x1 - 1;
        if ~all(img(:, x1) == col)
            break;
        end
    end
    % search from bottom for the first row that not contains only the border-color
    y0 = 0;
    while y0 < h
        y0 = y0 + 1;
        if ~all(img(y0, :) == col)
            break;
        end
    end
    % search from top for the first row that not contains only the border-color
    y1 = h + 1;
    while 1 < y1
        y1 = y1 - 1;
        if ~all(img(y1, :) == col)
            break;
        end
    end
    % calculate the sub image size to cut out
    x0 = max(0, x0 - new_bordersize);
    x1 = min(w, x1 + new_bordersize);
    y0 = max(0, y0 - new_bordersize - 7);
    y1 = min(h, y1 + new_bordersize);
    % return the new image
    img = img(y0 : y1, x0 : x1, :);
end

[DGM]

Trimming after the fact is not ideal, but it is often the only option left.

That said, the whole trimming process can be simplified. In this example, I use an image with a non-white background for sake of demonstration. While I'm reading the image from a file for simplicity, the same applies to an image as captured using getframe()/frame2im().

The original padding is removed and new padding added. In this case, the result demonstrates that the new padding is not constrained by the original amount of padding.

% inputs
inpict = imread('stuffthings.png');
padwidth = [50 0]; % [y x]

% create a mask of non-BG pixels
mask = ~all(inpict == inpict(1,1,:),3);

% find ROI geometry
mkr = any(mask,2);
mkc = any(mask,1);
r1 = find(mkr,1,'first');
r2 = find(mkr,1,'last');
c1 = find(mkc,1,'first');
c2 = find(mkc,1,'last');

% crop original image to extents
roipict = inpict(r1:r2,c1:c2,:,:);

% re-pad as specified
% doing this with IPT or base tools is clumsy, but it's fine
% doing the padding this way instead of shifting subscripts 
% allows the output padding to be unconstrained by the original image size
szo = size(roipict,1:2)+2*padwidth;
outpict = zeros(szo,class(roipict));
for c = 1:size(inpict,3)
    outpict(:,:,c) = padarray(roipict(:,:,c),padwidth,inpict(1,1,c),'both');
end

imshow(inpict,'border','tight')

figure
imshow(outpict,'border','tight')

Of course, there are more convenient ways. If I were doing it with MIMT, this simplifies to:

% inputs
inpict = imread('stuffthings.png');
padwidth = [100 10]; % [y x]

% create a mask of non-BG pixels
bg = inpict(1,1,:);
mask = ~all(inpict == bg,3);

% crop original image to extents
[~,rr,cc] = crop2box(mask);
outpict = inpict(rr,cc,:,:);

% re-pad as specified
outpict = addborder(outpict,padwidth,ctflop(bg));
imshow(outpict,'border','tight')

... or more simply:

% inputs
inpict = imread('stuffthings.png');
padwidth = [100 10]; % [y x]

% crop uniform border vectors
outpict = cropborder(inpict,NaN(1,4));

% re-pad as specified
bg = inpict(1,1,:);
outpict = addborder(outpict,padwidth,ctflop(bg));
imshow(outpict,'border','tight')

Answer 7

[Chris]

This is by no means the most elegant solution, but in my case I was saving figures as *.png files and then cropping them down once I'd inserted them onto a PowerPoint slide. After about 60 of these I found this thread. Since I'm the only one doing this work and I'm always working on the same monitor, I developed the following workflow:

1. create an image that is the size of my monitor,
2. save the figure as a PNG file,
3. load it back in using imread,
4. crop the image matrix down to the height and width I want using the associated row/column ranges,
5. save it out again overwriting the original PNG using imwrite.

The code looks like this:

test = rand(100,1);

f = figure('units','normalized');

plot(test,'LineWidth',3)

legend('Random Noise')

set(gca,'LineWidth',2,'YLim',[-0.1, 1.1],'FontSize',64)

set(f,'Position',[0,0.0333,1.0000,0.9132])

saveas(f,'test','png')

im = imread('test.png');

imwrite(im(100:2025,461:5000,:),'test.png')

The difference in the saved images goes from this:

->||<-

To this:

->||<-

I've been working on a large monitor and haven't noticed any negative impact on the image quality.