It is possible to compress images using singular value decomposition and leaving out the smaller values. You will get "squares" as artifacts.
Is something similar possible for a dataset of voxels?
Yes, you will get "cubes" as artifacts.
>>7807599
Ok. But how are they defined? How are they called? How can they be calculated?
Ok. Found the Tucker Decomposition.
What are you trying to use this for? I'm pretty interested in 3D computer vision as well