Question

the getsnapshot function takes a lot of time executing since (I guess) initializes the webcam every time is called. This is a problem if you want to acquire images with a high framerate.

I trick I casually discovered is to call the preview function, which keeps the webcam handler open making getsnapshot almost instantaneous, but it keeps a small preview window open:

% dummy example
cam = videoinput(...);
preview(cam);

while(1)
    img = getsnapshot(cam);
    % do stuff
end

Is there a “cleaner” way to speedup getsnapshot? (without preview window opened)

Answer 1

You could use the new “machine vision” toolbox which is specially built for vision applications. See code below:

vid = videoinput('winvideo', 1, 'RGB24_320x240'); %select input device

hvpc = vision.VideoPlayer;   %create video player object

src = getselectedsource(vid);
vid.FramesPerTrigger =1;
vid.TriggerRepeat = Inf;
vid.ReturnedColorspace = 'rgb';
src.FrameRate = '30';
start(vid)

%start main loop for image acquisition
for t=1:500
  imgO=getdata(vid,1,'uint8');    %get image from camera
  hvpc.step(imgO);    %see current image in player
end

As you can see, you can acquire the image with getdata. The bottleneck in video applications in Matlab was the preview window, which delayed to code substantially. The new vision.VideoPlayer is a lot faster (i have used this code in real time vision applications in Matlab. When i had written the first version without the vision toolbox, achieving frame rates at about 18 fps and using the new toolbox got to around 70!).

Note: I you need speed in image apps using Matlab, you should really consider using OpenCV libs through mex to get a decent performance in image manipulation.