matlab code not working for 64x64 dct block

I have implemented 2d dct code which is working for (4x4 ,8x8,16x16)block but its not working for 64x64 block ,in matlab its showing busy for more than 1 hour,what changes i should do to get the desired result.

for N = [8];
C = zeros(N,N);
for m = 0:1:N-1
for n = 0:1:N-1
if n == 0
k = sqrt(1/N);
else
k = sqrt(2/N);
end
C(m+1,n+1) = k*cos( ((2*m+1)*n*pi) / (2*N));
end
end
% Get Basis Functions
figure;
colormap('gray');
for m = 0:1:N-1
for n = 0:1:N-1
subplot(N,N,m*N+n+1);
Y = [zeros(m,N);
     zeros(1,n) 1 zeros(1,N-n-1);
    zeros(N-m-1,N)];
X = C*Y*C';
imagesc(X);
axis square;
axis off;
end
end
end

Please help i m trying a lot.

[Image Analyst]

What's the point of this for loop:

for N = [8];
    .....
end

You don't need a for loop to execute just one time with N=8. Just set N=8, and don't have a for loop.

[Joseph Cheng]

What i take out of the for N= [8] is that this person was looking to perform this section of code on for N=[2 4 8 16 ... 64]

Answer 1

[Walter Roberson]

The code finishes in a small number of seconds when I try it.

[pragyan bisoyi]

 @Walter Roberson my code is not working.

[Joseph Cheng]

you probably do not have enough memory in your system to generate 64x64 subplots. what you can do is preallocate and index them and put them together like below

figure;
colormap('gray');
X = zeros(64*64,64*64);
for m = 0:1:N-1
    for n = 0:1:N-1
        disp([m n])
        Y = zeros(N,N);
        Y(m+1,n+1) = 1;
        X(m*64+1:(m+1)*64,n*64+1:(n+1)*64) = C*Y*C';
        axis square;
        axis off;
    end
end
x = linspace(0,64,64*64);
imagesc(x,x,X),colormap(gray),axis equal, axis tight, grid on
set(gca,'ytick',0:64)
set(gca,'xtick',0:64)

also i simplified what you were trying to do when creating Y.

[Walter Roberson]

I copied your exact code and it worked fine on my R2014a system. It took about 0.8 seconds to execute.

[Joseph Cheng]

Walter the code above works fine but the problem is going up and up for N=64 such that the block of data is 64x64. the code above is the 8x8 case (line 1)

[Walter Roberson]

Ah yes, it does take a long time to build the 4096 individual plots.

The C matrix generation could obviously be optimized some, but that probably does not matter compared to the plot time.

[mg, ng] = ngrid(0:1:N-1);

k = sqrt(2/N);
C = k .* cos( ((2*mg+1) .* ng .* pi) ./ (2*N));
C(:,1) = C(:,1) ./ sqrt(2);

That is, build it all using the larger k, and then correct the first column which is now sqrt(2) too large. k*cos( ((2*m+1)*n*pi) / (2*N));