Scale up an image without interpolation

I have 3D image, which I need to scale up with integer scaling coefficient (e.g. x3, x5). ImageJ have an interpolation option "None", alongside with "Bilinear" and "Bicubic", but Matlab functions (like resize, imresize) do not include "none" option. It is useful for overlaying high resolution image over original low resolution one. I need to replace each voxel with a volume of identical value pixels. Can I do something smarter then nested FOR loops?

Answer 1

[Image Analyst]

The option is not called 'none', it's called 'nearest'. This will basically replicate voxels, not interpolate them. I'm sure your ImageJ is doing it the same way.

bigArray = imresize(smallArray, 3.456, 'nearest');

[Valentin Magidson]

Thanks, it works great for 2D images. For 3D it operates on each 2D slice separately, but repmat(Matrix2D,1,1,scale) can expand each slice.

[Patryk Trybalski]

Ain't 'nearset' the nearest neightbor interpolation?

Answer 2

[Adam]

Something like this should work I think, with a single for loop. I'm not sure if it could be done without a for loop at all. Probably some combination of repmat and reshape may work to do the whole thing at once rather than a row at a time.

scaleFactor = 3;
inputImage = rand( 3, 4 );
outputImage = zeros( size( inputImage ) * scaleFactor );
nOutputColumns = size( outputImage, 2 );
outputRowIndices = 1:scaleFactor:size( outputImage, 1 );

for i = 1:size( inputImage, 1 )
   idx = outputRowIndices(i);
   outputImage( idx:idx+2, : ) = reshape( repmat( inputImage(i,:), scaleFactor^2 , 1 ), scaleFactor, nOutputColumns );
end