Similar Tensor Arrays – a Framework for Storage of Tensor Data

This paper briefly describes a framework for storage of geometric tensor array data, useful for storage of regularly sampled tensor fields and regularly sampled tensor-valued functions on charts of manifolds in differential geometry. The framework, called Similar Tensor Array Core Headers, abbreviated STACH and pronounced like the english word ”stash”, capture the essence of tensor field processing in a minimalistic set of attributes. It can be used as a “greatest common divisor” in tensor processing algorithms and guide users in applied fields such as medical image analysis, visualization and manifold learning, to store and handle tensor array data in a standardized way. The framework solves many problems for new users of tensor data, promote a mathematical and geometric view of tensors and encourage exchange of tensor data between different labs and different fields of research.