Designing a Software Framework for Multimodal Hyperspectral Image Analysis in Painting Conservation

Multimodal imaging, encompassing hyperspectral imaging (HSI), macro X-ray fluorescence (MA-XRF), and RGB photography, is essential for the scientific investigation of paintings. However, analysis workflows remain fragmented across specialised tools, scripting environments, and image editors, complicating cross-modal reasoning and constraining integrated analysis. To address this, this paper introduces a software concept for investigating structured integration strategies for multimodal painting analysis tools. The concept is implemented in SpectraPaint , a prototype web application under active development. The purpose of this work is to establish the foundations for such an environment: an extensible core infrastructure, a conceptual data model adequate to multimodal painting analysis focusing on HSI, and a prototype instantiation that demonstrates both. The architecture follows a microkernel-style design that separates stable infrastructure from extensible analytical components, while the domain-informed conceptual data model maps the data, metadata, and paradata of multimodal painting analysis onto established interoperability standards, e.g., Dublin Core or FAIR practices. The data model is structured around a configurable Extract-Transform-Load (ETL) pipeline that links spectral extraction and classification as reproducible steps. By investigating these elements, the work also identifies and explores central technical requirements for this type of application. The technical decisions regarding the storage and deployment are geared toward a local prototype that enables rapid development; the data model, however, is not bound to this setting and remains valid beyond it, including for a potential cloud-based deployment. The framework is demonstrated through a case study on Edvard Munch's Old Man in Warnemünde , illustrating how multimodal datasets can be integrated into a unified exploratory environment supporting iterative spectral extraction and region-of-interest (ROI)-based comparison. SpectraPaint serves as a reusable architectural foundation for systems bridging computational imaging and interdisciplinary conservation practice.