Efficient manipulation of images in memory

Which is the more efficient, to store and manipulate several images in an image cube (M,N,K) or in a cell array of images cell(K,1)? I'm thinking of memory, but other measures of efficiency are welcome.

Answer 1

[Jan]

The answer depends on the values of M*N and K. If K is very large, reserving a single [M*N*K] memory block is expensive and a {K,1} cell of [M*N] blocks should be preferred. For small K, the single memory block can be more efficient.

If M*N is very large, extracting a single picture from the [M*N*K] memory block is expensive, because a data copy is necessary. Then a cell vector can be more efficient - but this depends on the actually procedure applied to the images.

My conclusion: Both of your ideas are good and have benefits. Implement both and share your results here.

[Joe]

Thanks, Jan. I usually use cubes unless I need to change the size of images. Cells make it easier to do that without creating entirely new cubes or arrays. I knew there must be more behind the choice than that.

Answer 2

[Sean de Wolski]

If the images are all the same size, use the cube. If not, you would have to pad the smaller images to the size of the biggest which wastes memory and a cell would be better.

Cubes are better for manipulations. my $0.02