N-dimensional Fourier interpolation

Performs N-D FFT interpolation with upsampling, downsampling, or mixed up- and downsampling

Description

Performs N-D FFT interpolation on any data for which fftn works. Will upsample by zero-filling, downsample by truncating high frequencies, or combine both up- and downsampling by dimension to allow arbitrary reshaping.

• MATLAB release: MATLAB 7 (R14)

[Matthias Schabel]

This code is needed as well (append on to end of fftInterpolate.m file):

function subData = subRange(data,rng)

if (isempty(rng))
    subData = data;
    return;
end;

sz = size(data);
dim = length(sz);

lo = rng(:,1)';
hi = rng(:,2)';

if (length(lo) ~= dim || length(hi) ~= dim)
    error('subRange :: dimension mismatch');
end;

% replace zeros with lower/upper limit
for i=1:dim
    if (lo(i) == 0) lo(i) = 1; end;
    if (hi(i) == 0) hi(i) = sz(i); end;
end;

if (any(lo<1) | any(hi > sz))
    error('subRange :: sub-range out of bounds');
end;

switch (dim)
    case 1, subData = data(lo(1):hi(1));
    case 2, subData = data(lo(1):hi(1),...
                    lo(2):hi(2));
    case 3, subData = data(lo(1):hi(1),...
                    lo(2):hi(2),...
                    lo(3):hi(3));
    case 4, subData = data(lo(1):hi(1),...
                    lo(2):hi(2),...
                    lo(3):hi(3),...
                    lo(4):hi(4));
    case 5, subData = data(lo(1):hi(1),...
                    lo(2):hi(2),...[]
                    lo(3):hi(3),...
                    lo(4):hi(4),...
                    lo(5):hi(5));
    otherwise,
            % generate string and use eval
            str = 'subData = data(';
            for i=1:dim
                str = [str 'lo(' num2str(i) '):hi(' num2str(i) ')'];
                if (i~=dim)
                    str = [str ','];
                else
                    str = [str ');'];
                end;
            end;
            eval(str);
end;
        
return;

[Benjamin]

Matthias, I cannot get this to work. The example image posted seems to interpolate the peaks dataset by a factor of two. But when I run the following code, I don't get the same result:
>>> figure; subplot(1,2,1); imagesc(peaks);
>>> subplot(1,2,2); imagesc(fftInterpolate(peaks,2));
what's going on?

[Matthias]

code works excellent for data with even dimensions. In case of odd dimensions the code fails.

[Matthias]

Dear author

please find
rng = [ceil(newsz/2)-ceil(sz/2)+1;ceil(newsz/2)-ceil(sz/2)+sz]';

and replace it with
rng = [ceil(newsz/2)-floor(sz/2)+1;ceil(newsz/2)-floor(sz/2)+sz]';

then the code also works with odd dimesions.

Matthias Wurm

[Kevin Sharp]

This fails with ...

??? Undefined function or method 'odd' for input arguments of type 'double'.

Error in ==> fftInterpolate at 61
centeru = floor(newsz/2)+1+odd(newsz);