Render MATLAB figure in memory

*爱你&永不变心* 提交于 2019-11-26 04:46:26

问题


Are there any alternatives to using getframe and saveas for saving the contents of a figure to a raster image for further processing?

Approach 1: getframe

h = figure(\'visible\', \'off\');
a = axes(\'parent\', h);

% render using `scatter3()` or other plot function.

content = frame2im(getframe(h));

This has the serious drawback of showing the figure to perform a screen capture in the call to getframe() and it is problematic when performing such a render in a loop (i.e. saving content at each iteration as a video frame).

Approach 2: saveas

h = figure(\'visible\', \'off\');
a = axes(\'parent\', h);

% render using `scatter3()` or other plot function.

saveas(h, \'/path/to/file.png\');
content = imread(/path/to/file.png\');

This approach has the serious drawback of writing to disk, which is problematic in multithreaded applications, as well as being slower than rendering directly to memory. Since saveas() will obviously render to memory before invoking the PNG encoder, what I want is possible, but I can\'t find any function it in the MATLAB documentation that only performs the rendering part.

Question:

Does you know of an alternate way of rendering an arbitrary axes content to a raster image?


回答1:


If you create an avi file with avifile, and then add frames to it with addframe, MATLAB doesn't open up extra visible figures like it does with getframe.

avi = avifile('/path/to/output');
figure_handle = figure('visible', 'off');

% ...
for something = 1:1000
    cla
    % (draw stuff...)
    avi = addframe(avi, figure_handle);
end



回答2:


I realize this is an old thread, but I ran into this problem again lately, so I wanted to summarize my findings. My main source is this page (cached). According to it, there are three alternatives:

  1. using ADDFRAME directly with the figure handle (without using GETFRAME). This is exactly what @rescdsk has shown in his answer.

    hFig = figure('Visible','off');
    
    aviobj = avifile('file.avi');
    for k=1:N
        %#plot(...)
        aviobj = addframe(aviobj, hFig);
    end
    aviobj = close(aviobj);
    
  2. using PRINT/SAVEAS/HGEXPORT to export the figure to an image file, then reading the image back from disk. This is approach#2 that you listed yourself in the question above.

    hFig = figure('Visible','off');
    set(hFig, 'PaperPositionMode','auto', 'InvertHardCopy','off')
    
    aviobj = avifile('file.avi');
    for k=1:N
        %#plot(...)
        print(['-f' num2str(hFig)], '-zbuffer', '-r0', '-dpng', 'file.png')
        img = imread('file.png');
        aviobj = addframe(aviobj, im2frame(img));
    end
    aviobj = close(aviobj);
    
  3. using the undocumented HARDCOPY function to capture the figure in-memory.

    hFig = figure('Visible','off');
    set(hFig, 'PaperPositionMode','auto')
    
    aviobj = avifile('file.avi');
    for k=1:N
        %#plot(...)
        img = hardcopy(hFig, '-dzbuffer', '-r0');
        aviobj = addframe(aviobj, im2frame(img));
    end
    aviobj = close(aviobj);
    

    In fact, this is the underlying function that other functions use either directly or indirectly. By inspecting the source codes where possible, here is an illustration of the dependencies of the related functions where A --> B denotes A calls B:

    saveas [M-file] --> print [M-file] --> render [private M-file] --> hardcopy [P-file]
    hgexport [P-file] --> print [M-file] --> ...
    @avifile/addframe [M-file] --> hardcopy [P-file]
    

    On the other hand, GETFRAME does not call HARDCOPY but an undocumented built-in function named CAPTURESCREEN (although it seems that it will be using PRINT for the upcoming HG2 system where there is a new -RGBImage print flag):

    getframe [M-file] --> capturescreen [builtin]
    

Note: Since AVIFILE is now deprecated, you can replace it with the newer VIDEOWRITER in (2) and (3), but not in (1) since it does not support passing figure handle directly.




回答3:


Start MATLAB in headless mode: matlab -noFigureWindows

MATLAB is running in headless mode. Figure windows will not be displayed.

then simply plot and save the figures as usual (you won't see any graphical output of course). Example:

surf(peaks);
print output.eps     %# SAVEAS works as well
close

I tested the above on a Windows machine running R2010a. I don't have access to a Unix machine right now, but I answered a similar question in the past, and it worked just fine at the time (you will need to unset the $DISPLAY variable before starting MATLAB)


EDIT

Another option, in case you want to keep your normal workspace, is to start a new MATLAB instance in the background which will generate and save the plots (source).

Run this from the command prompt of your current MATLAB session (all on the same line):

!start /B /MIN matlab -noFigureWindows 
                      -automation 
                      -r "cd('c:\yourpath'); myscript; quit"

This will start a new MATLAB session in the background (using COM Automation), and execute a script called myscript (a simple M-file) that contains all your plotting code:

c:\yourpath\myscript.m

surf(peaks);
saveas(gcf, 'output.eps');



回答4:


With avifile being deprecated, this is how you do it with VideoWriter:

hFig = figure('Visible','off');
set(hFig, 'PaperPositionMode','auto')

aviobj = VideoWriter('file','Archival');
for k=1:N
    %#plot(...)
    img = hardcopy(hFig, '-dzbuffer', '-r0');
    writeVideo(aviobj, im2frame(img));
end
close(aviobj);


来源:https://stackoverflow.com/questions/4137628/render-matlab-figure-in-memory

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!