comparison DL/RLS-DLA/private/countcover.m @ 60:ad36f80e2ccf

(none)
author idamnjanovic
date Tue, 15 Mar 2011 12:20:59 +0000
parents
children
comparison
equal deleted inserted replaced
59:23f9dd7b9d78 60:ad36f80e2ccf
1 function cnt = countcover(sz,blocksize,stepsize)
2 %COUNTCOVER Covering of signal samples by blocks
3 % CNT = COUNTCOVER(SZ,BLOCKSIZE,STEPSIZE) assumes a p-dimensional signal
4 % of size SZ=[N1 N2 ... Np] covered by (possibly overlapping) blocks of
5 % size BLOCKSIZE=[M1 M2 ... Mp]. The blocks start at position (1,1,..,1)
6 % and are shifted between them by steps of size STEPSIZE=[S1 S2 ... Sp].
7 % COUNTCOVER returns a matrix the same size as the signal, containing in
8 % each entry the number of blocks covering that sample.
9 %
10 % See also IM2COLSTEP, COL2IMSTEP, IM2COL.
11
12
13 % Ron Rubinstein
14 % Computer Science Department
15 % Technion, Haifa 32000 Israel
16 % ronrubin@cs
17 %
18 % August 2008
19
20
21 cnt = ones(sz);
22 for k = 1:length(sz)
23
24 % this code is modified from function NDGRID, so it computes one
25 % output argument of NDGRID at a time (to conserve memory)
26 ids = (1:sz(k))';
27 s = sz; s(k) = [];
28 ids = reshape(ids(:,ones(1,prod(s))),[length(ids) s]);
29 ids = permute(ids,[2:k 1 k+1:length(sz)]);
30
31 cnt = cnt .* max( min(floor((ids-1)/stepsize(k)),floor((sz(k)-blocksize(k))/stepsize(k))) - ...
32 max(ceil((ids-blocksize(k))/stepsize(k)),0) + 1 , 0 );
33 end