Bernd

the function only threw error messages at me, but looking at the code, I foun what I needed so badly; the extraction of image data in a subplot

w=findobj(gcf,'Type','image');
ww=get(w,'cdata');

stores cdata of every image into a cell of ww

The convolution is very fast and pretty accurate for the 'valid' part of an 2D signal (except the known double-single precision difference), but there are big differences near the edges if using 'same' shape. Therefore I wrote a piece of shaping code to treat it like conv2. Please test and report any coding mistakes!!!
____________________________________________________
function [newimage] = cudaconv2(image,filter,shape)
if nargin == 2
    shape = 'full';
end

if (strcmp(shape, 'full')) % it's not a real 'full' convolution !!!!!
    [im in] = size(image);
    [fm fn] = size(filter);
    outM1 = 1;
    outN1 = 1;
    image2 = zeros(im+fm-1, in+fn-1);
    image2(round(fm/2):round(im + fm/2 - ...1),round(fn/2):round(in + fn/2 - 1)) = image(1:end,1:end);
    output = cudaconv(image2,filter);
    [outM2, outN2] = size(output);
    
elseif (strcmp(shape, 'same')) % large differences on the edges
    output = cudaconv(image,filter);
    [Am An] = size(image);
    outM1 = 1;
    outN1 = 1;
    outM2 = Am;
    outN2 = An;

elseif (strcmp(shape, 'valid')) % very accurate
    output = cudaconv(image,filter);
    [Am An] = size(image);
    [Cm Rn] = size(filter);
    outM1 = round(Cm/2);
    outN1 = round(Rn/2);
    outM2 = round(Am - Cm/2);
    outN2 = round(An - Rn/2);
else
    disp('Shape type not valid');
    return;
end

newimage = output(outM1:outM2,outN1:outN2);
____________________________________________________