wt_dim_shifts

 return dimension permute shifts needed for wt/iwt routines
 FORMAT [shifts, end_shift] = wt_dim_shifts(p_dims)
 
 Inputs
 
 p_dims      - flags, one for each dimension, non-zero if dimension is
               to be (wt, iwt) processed
 sz          - size of matrix to be processed
               (needed to identify row vectors)
 
 Outputs 
 
 shifts      - values, one for each non-zero value of p_dims, giving how
               many extra dimensions to shift to get the current (p_dim)
               dimension to be X (columns)
   
 end_shift   - how many shifts right to get from last processed
               dimension back to original matrix orientation

 For example, if processing all 3 dimensions of a 3D matrix, we will need
 to specify 4 shifts.  When processing the first (X) dimension, we do not
 have to do a shift (shift(1) = 0).  For Y we will need to shift Y to X
 (shift(2) = -1).  For Z, we further need to shift Z to X: shift(3) = -1;
 Last, to shift back to where we started (X back to X) we need an end shift
 (end_shift = -1).  This will result in the following permutes, in the
 wt/iwt routines:
 
 x1 = permute(x,  [1 2 3]);  % shift(1) of 0 
 x2 = permute(x1, [2 3 1]);  % Y->X; shift(2) of -1
 x3 = permute(x2, [2 3 1]);  % Z->X; shift(3) of -1
 x  = permute(x3, [2 3 1]);  % X->X; end_shift of -1
 
 $Id: wt_dim_shifts.m,v 1.1 2004/09/26 04:00:24 matthewbrett Exp $

0001 function [shifts, end_shift] = wt_dim_shifts(p_dims, sz)
0002 % return dimension permute shifts needed for wt/iwt routines
0003 % FORMAT [shifts, end_shift] = wt_dim_shifts(p_dims)
0004 %
0005 % Inputs
0006 %
0007 % p_dims      - flags, one for each dimension, non-zero if dimension is
0008 %               to be (wt, iwt) processed
0009 % sz          - size of matrix to be processed
0010 %               (needed to identify row vectors)
0011 %
0012 % Outputs
0013 %
0014 % shifts      - values, one for each non-zero value of p_dims, giving how
0015 %               many extra dimensions to shift to get the current (p_dim)
0016 %               dimension to be X (columns)
0017 %
0018 % end_shift   - how many shifts right to get from last processed
0019 %               dimension back to original matrix orientation
0020 %
0021 % For example, if processing all 3 dimensions of a 3D matrix, we will need
0022 % to specify 4 shifts.  When processing the first (X) dimension, we do not
0023 % have to do a shift (shift(1) = 0).  For Y we will need to shift Y to X
0024 % (shift(2) = -1).  For Z, we further need to shift Z to X: shift(3) = -1;
0025 % Last, to shift back to where we started (X back to X) we need an end shift
0026 % (end_shift = -1).  This will result in the following permutes, in the
0027 % wt/iwt routines:
0028 %
0029 % x1 = permute(x,  [1 2 3]);  % shift(1) of 0
0030 % x2 = permute(x1, [2 3 1]);  % Y->X; shift(2) of -1
0031 % x3 = permute(x2, [2 3 1]);  % Z->X; shift(3) of -1
0032 % x  = permute(x3, [2 3 1]);  % X->X; end_shift of -1
0033 %
0034 % $Id: wt_dim_shifts.m,v 1.1 2004/09/26 04:00:24 matthewbrett Exp $
0035 
0036 n_dims = length(p_dims);
0037 fpd  = find(p_dims);
0038 
0039 % First check for situations where we do not have to do any shifts
0040 if all(fpd == 1) | ...           % only processing X dimension
0041   (n_dims == 2 & sz(1) == 1)     % row vector
0042   shifts =  zeros(1, sum(p_dims));
0043   end_shift = 0;
0044 else  % we do have to do shifts
0045   shifts = diff([1 fpd n_dims+1]) * -1;
0046   end_shift = shifts(end);
0047   shifts = shifts(1:n_dims);
0048 end
0049