Deep Hierarchical Super Resolution for Scientific Data

TL;DR

This paper introduces a hierarchical super resolution method for volumetric scientific data that reduces artifacts and supports varying levels of detail, outperforming traditional approaches.

Contribution

It proposes a novel hierarchical neural network approach for super resolution of octree-structured data, enabling flexible, high-quality upscaling across different levels of detail.

Findings

Outperforms baseline interpolation and hierarchical upscaling methods

Reduces seam artifacts at octree node boundaries

Demonstrates effectiveness in data reduction and visualization tasks

Abstract

We present a novel technique for hierarchical super resolution (SR) with neural networks (NNs), which upscales volumetric data represented with an octree data structure to a high-resolution uniform grid with minimal seam artifacts on octree node boundaries. Our method uses existing state-of-the-art SR models and adds flexibility to upscale input data with varying levels of detail across the domain, instead of only uniform grid data that are supported in previous approaches. The key is to use a hierarchy of SR NNs, each trained to perform 2x SR between two levels of detail, with a hierarchical SR algorithm that minimizes seam artifacts by starting from the coarsest level of detail and working up. We show that our hierarchical approach outperforms baseline interpolation and hierarchical upscaling methods, and demonstrate the usefulness of our proposed approach across three use cases including data reduction using hierarchical downsampling+SR instead of uniform downsampling+SR, computation savings for hierarchical finite-time Lyapunov exponent field calculation, and super-resolving low-resolution simulation results for a high-resolution approximation visualization.