How to create subplots with little vertical spacing?

Hello

I want to do a figure in Matlab consisting of a grid of images (subplots). I have used the subaxis toolbox :

subaxis(4,6,1, 'Spacing', 0.03, 'Padding', 0, 'Margin', 0);

The problem is that the vertical space between the subplots (images) is too big.

How can this be solved? The horizontal spacing is ok but I want the vertical spacing the same as the horizontal spacing. With the above toolbox this seems not working.

I'm also interested in other solutions than the mentioned toolbox.

[KAE]

You could try tiledlayout.

Answer 1

[Kelly Kearney]

The subaxis command allows you to specify different values for vertical and horizontal spacing. Try playing around with different values to get what you want. Remember that the spacing/padding/margins are defined in terms of normalized coordinates, so if your figure isn't square, the vertical and horizontal distances won't be quite the same.

For example, the following will give you very little vertical spacing:

iax = 1; % Or whichever
subaxis(4, 6, iax, 'sh', 0.03, 'sv', 0.01, 'padding', 0, 'margin', 0);

[Sepp]

It did not decrease the vertical spacing. I think the problem is that my images are not square. The dimensions are 85 (width) x 35 (height).

[Kelly Kearney]

Are you by any chance plotting to these axes using imshow, or calling axis equal or some other equivalent? In that case, you need to understand the difference between the axis position and the plot box position. The plot box is modified to fit your image, but the axis position stays the same:

figure('color', 'w');
props = {'spacing', 0.03, 'margin', 0};
ax(1) = subaxis(4,6,1, props{:});
ax(2) = subaxis(4,6,2, props{:});
ax(3) = subaxis(4,6,7, props{:});

for ii = 1:3
    imagesc(rand(35,85), 'parent', ax(ii));
end

arrayfun(@(ax) axis(ax, 'equal'), ax)

arrayfun(@(ax) axis(ax, 'tight'), ax)

So while the spacing between the axes is exactly as you've specified, there's a lot of extra space introduced by the axis rescaling.

The best way to make sure your axes take up exactly as much space as you need would be to position them manually, as Cedric suggested. Or you could calculate the exact margins and spacing you would need in order to get all the subplots where you want them (mostly by adding extra margin space above and below).

[Sepp]

I have found the problem. It is the axis('image'); command. Here is what I'm doing:

figure('Name', 'Brain Slices', 'NumberTitle', 'off', 'units', 'normalized', 'outerposition', [0 0 1 1]);

for i = 1:size(rawVolume,3)
    slice = rawVolume(:,:,i,1);
    slice= squeeze(slice);
    slice = rot90(slice);
    subaxis(3,4,i, 'Spacing', 0.03, 'Padding', 0, 'Margin', 0);
    imagesc(slice);
    colormap gray;
    caxis([ 0 2300 ]);
    axis('image');
    axis('tight');
    axis('off');
end

With axis('image'); the results looks as follows:

Without axis('image'); it looks as follows:

The problem is now that the images are stretched in y-direction. How can I get this spacing without any stretching?

[Kelly Kearney]

As I said, you have to do the calculations... Assuming you want 3% spacing horizontally, and the same spacing vertically (in pixels, not percentage):

rawVolume = rand(85,35,24);

nr = 6;    % # rows
nc = 4;    % # columns
sh = 0.03; % horizontal spacing (normalized units)

f = figure;
fpos = get(f, 'position');

% Convert 0.03 normalized to pixels by multiplying
% by figure width

shpx = fpos(3)*sh;  

% Calculate subplot width, in pixels
% (Figure width-spacing between subplots)/# plots

wpx = (fpos(3) - shpx*(nc-1))/nc;

% Calculate subplot height: proportional to width
% based on image size

hpx = (size(rawVolume,2)./size(rawVolume,1)) * wpx;

% You want the same spacing vertically as horizontally,
% translate from pixels to normalized units by diving
% by figure height

sv = shpx./fpos(4);

% The leftover vertical space (figure height - vertical
% spacing between subplots - height of 4 subplots) will
% be the bottom margin. Divide by figure height to get
% normalized units.

mar = (fpos(4) - hpx*nr - shpx*(nr-1))./fpos(4);

% Now plot

props = {'sh', sh, 'sv', sv, 'mar', 0, 'mb', mar};
for ii = 1:size(rawVolume,3)
    ax(ii) = subaxis(nr,nc,ii, props{:});
    imagesc(rot90(rawVolume(:,:,ii)));
    axis('image', 'tight', 'off');
end

[Sepp]

Thank you for your detailed answer. I will try it out. Would it be possible that you can comment your script? I just don't understand the calculations.

[Sepp]

I have tried it out. It works perfectly. Thank you so much for it.

[Kelly Kearney]

I've added a bunch of comments to the example above to walk you through exactly what each line does.

Answer 2

[Cedric]

You can store/use the axis handles of both subplots and access/modify their properties. Here is an example:

 t = 0:0.1:10 ; 

 hAxis(1) = subplot( 2, 1, 1 ) ; 
 plot( t, sin(t) ) ;

 hAxis(2) = subplot( 2, 1, 2 ) ;
 plot( t, cos(t) ) ;

This produces:

Then you can access axes properties through their handles. E.g. for querying all available properties:

 >> set( hAxis(1) )
                       ALim: {}
                   ALimMode: {'auto'  'manual'}
     ActivePositionProperty: {'position'  'outerposition'}
     ...

Browsing all properties, you realize that the Position property is probably what you need to update. For getting a specific property, e.g. Position:

 >> pos = get( hAxis(1), 'Position' )
 pos =
    0.1300    0.5838    0.7750    0.3412

where the first two coordinates are the horizontal and vertical relative (in [0,1]) coordinates of the origin of the axis with respect to the lower left corner of the figure, and the second two are the relative width and height. Say you want to make the first axis taller and closer to the second:

 >> pos(2) = 0.5 ;                         % Shift down.
 >> pos(4) = 0.45 ;                        % Increase height.
 >> set( hAxis(1), 'Position', pos ) ;

and when you execute it, you get:

PS: I would favor this type of approache over 3rd party functions, because it is easy to learn using handles, and then you are really free to design almost whatever you want (e.g. small inserts in larger plots). For this purpose, the last thing that you need to know is the AXES function, which creates a new axes graphic object in a figure. To illustrate:

 hAxis(3) = axes( 'Position', [0.66, 0.55, 0.2, 0.2] ) ;
 [X,Y,Z] = peaks( 15 ) ;
 surf( X, Y, Z);

adds the following insert:

[Cedric]

Last example, for the fun of it:

 load clown ;

 nRows = 3 ;
 nCols = 5 ;

 % - Create figure, set position/size to almost full screen.
 figure() ;
 set( gcf, 'Units', 'normalized', 'Position', [0.1,0.1,0.8,0.8] ) ;

 % - Create grid of axes.
 [blx, bly] = meshgrid( 0.05:0.9/nCols:0.9, 0.05:0.9/nRows:0.9 ) ;
 hAxes = arrayfun( @(x,y) axes( 'Position', [x, y, 0.9*0.9/nCols, 0.9*0.9/nRows] ), blx, bly, 'UniformOutput', false ) ;

 % - "Plot data", update axes parameters.
 for k = 1 : numel( hAxes )
    axes( hAxes{k} ) ;
    image( X ) ;
    set( gca, 'Visible', 'off' ) ;
 end

 colormap( map ) ;

Here you see that it takes 2 lines of code for defining your own grid of axes. It is easy to wrap this into a function of your own, which fills the axes with whatever you need to plot/display.

[William Chamberlain]

jump to use subaxis if you just need to get it done right now (which is generally when I'm Googling for answers), but definitely come back and learn this ASAP, because this is a better solution than subaxis : you can easily set things up literally any way you want, and I'm kicking myself for not learning this a year ago.

Answer 3

[Walter Roberson]

subaxis(4,6,1, 'Spacing', 0.03, 'Padding', 0, 'Margin', 0, 'SpacingVert', 0.03);