how can i solve this problem about 'OUT OF MEMORY'

6 次查看（过去 30 天）
Beyung Doo Jo
Beyung Doo Jo 2011-8-9
when i use matrix, i always got out of memory.... actually i put zeros fuction to avoid it..
but, it is not work at all.
i'm using window 7 32bit 2G RAM Matlab 7.11.0 R2010b
Matrix size is 256 by 192
it's not big CT Image
Even if it takes a long tome, i think that it has to be worked.
Someone told me that the latest version of matlab has problem on it..
how can i get out of this problem. plz help me..
i'm not good at Matlab.
My E-mail is misonara_@hotmail.co.kr
nx = length(x);
ny = length(y);
nz = length(z);
% Remark: There may be some savings in fft computation if P is transposed
[nv,nu] = size(P);
u = (-u_off + [0:nu-1]')*du;
v = (-v_off + [0:nv-1]')*dv;
% Remark: if P is transposed, 'meshgrid' should be changed to 'ndgrid'
[uu,vv] = meshgrid(u,v);
weight = SDD./sqrt(SDD^2 + vv.^2 + uu.^2);
% Remark: P is over-written to conserve memory in following computations.
% The interpretation at each stage should be clear from context
P = P .* weight;
% Filtering:
Phihat = varargin{1};
if ~isempty(Phihat), % Skip to backprojection is filter is empty
% Remark: This call to fft can likely be optimised using the transpose
% of P instead to capitalise on Matlab's column-oriented storage.
P = fft( P, length(Phihat), 2 ); % Rows zero-padded automatically
% Remark: This loop can likely be vectorised using repmat
for j=1:nv
P(j,:) = P(j,:) .* Phihat ;
end;
% Remark: This call to fft can likely be optimised using the transpose
% of P instead to capitalise on Matlab's column-oriented storage.
P = real( ifft( P, [], 2 ) );
P = P(:,1:nu); % Trim zero-padding on rows
end % of if-block
% Increment to add to reconstruction in backprojection stage
dR = zeros(ny,nx,nz);
% Vectorised computation of backprojection
[yy, xx, zz] = ndgrid( y, x, z);
% Use projection matrix to project reconstruction (x,y,z) grid
% into detector (u,v) grid
*UV = A * [ xx(:)'; yy(:)'; zz(:)'; ones(1,nx*ny*nz) ];*
* ---------->this part*
% Nearest neighbour interpolation to find detector coordinates (u,v)
% that are closest to projections of voxels (x,y,z)
U = round( UV(1,:)./UV(3,:) ) + 1 ;
V = round( UV(2,:)./UV(3,:) ) + 1 ;
% Identify indices of voxels with projections strictly within detector grid
ind = find( (U>=1) & (V>=1) & (U<=nu) & (V<=nv) );
% Remark: This will have to be changed if P is transposed
P_ind = ( U(ind) - 1 )*nv + V(ind) ;
  3 个评论
显示 1更早的评论
Oleg Komarov
Oleg Komarov 2011-8-9
Also add the error message and which line it refers to.
Beyung Doo Jo
Beyung Doo Jo 2011-8-11
*UV = A * [ xx(:)'; yy(:)'; zz(:)'; ones(1,nx*ny*nz) ];*
* ---------->this part*
just they said that i got ' out of memory ', when i using above code.

请先登录，再进行评论。

请先登录，再回答此问题。

回答（2 个）

Nirmal Gunaseelan
There is an excellent resource on understanding how MATLAB uses memory and how you can avoid out of memory issues - Avoiding Out of Memory issues - that might help you resolve such issues easily.
Sean de Wolski
Sean de Wolski 2011-8-9
Buy more RAM and upgrade to a 64bit system. CT scans, by nature, require a lot of memory. Also, you could convert to integers to save a little bit more.

类别

在 Help Center 和 File Exchange 中查找有关 Matrices and Arrays 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by