Embedding Meta Information into Visualizations

In this work, we study how to co-locate meta information with visualizations by directly embedding information into visualizations. This allows for visualizations to carry provenance and authorship information themselves for reproducibility. We call these self-describing visualizations-reproducible, authenticatable, and documentable. Self-describing visualizations can be used to extend existing visualization provenance systems. Herein, we start with a survey of existing digital image watermarking literature. We search for and classify watermarking algorithms that can support scientific visualizations. Using our payload-resilience testing framework, we evaluate and recommend algorithms supporting various use cases in the payload-resiliency space, and present guidelines for optimizing visualizations to improve payload capacities and embedding robustness. We demonstrate the efficacy of self-describing visualizations with two sample application implementations: (1) adding an embedding filter as a part the standard rendering pipeline, (2) creating a web reader to automatically and reliably extract provenance information from scientific publications for review and dissemination.