localSum

 Fast routine for box filtering.

 Same effect as calling 'C=convn( I, ones(dims), shape)', except more
 efficient. Computes local sums by using running sums. To get sum in
 non-overlapping windows, use shape='block'. Equivalent to doing localSum,
 and then subsampling (except more efficient). If operation op is set to
 'max' or 'min', computes local maxes or mins instead of sums. Note, that
 when applicable convSum and convMax are significantly faster.

 USAGE
  I = localSum( I, dims, [shape], [op] )

 INPUTS
  I       - matrix to compute sum over
  dims    - size of volume to compute sum over, can be scalar
  shape   - ['full'] 'valid', 'full', 'same', or 'block'
  op      - ['sum'] 'max', or 'min'

 OUTPUTS
  C       - matrix of sums

 EXAMPLE - 1
  A=rand(500,500,1); dim=25; f=ones(dim,1); shape='same'; r=20;
  tic, for i=1:r, B = localSum(A,dim,shape); end; toc
  tic, for i=1:r, C = conv2(conv2(A,f,shape),f',shape); end; toc
  diff=B-C; im(diff), sum(abs(diff(:)))

 EXAMPLE - 2
  load trees; I=ind2gray(X,map); figure(1); im(I);
  I1=localSum(I,3,'block','sum'); figure(2); im(I1); title('sum')
  I2=localSum(I,3,'block','max'); figure(3); im(I2); title('max')
  I3=localSum(I,3,'block','min'); figure(4); im(I3); title('min')

 See also convSum, convMax, imShrink

 Piotr's Image&Video Toolbox      Version 3.00
 Copyright 2012 Piotr Dollar.  [pdollar-at-caltech.edu]
 Please email me if you find bugs, or have suggestions or questions!
 Licensed under the Simplified BSD License [see external/bsd.txt]