Abstract
Cultural heritage imaging has specific needs with regards to the analysis of images that require the manipulation of a single digital object that combines the images obtained from different instruments probing different scales at different wavelengths, with the further possibility of selecting two or three dimensional representations. We propose a unified imaging data processing approach based on the Chebyshev Technology using the open source software Chebfun which, by mapping data processing to simple polynomial transformations, brought considerable improvements over already existing procedures. Within that same data processing framework we may further investigate how to merge images originating from different acquisition devices since all images are expressed in the same basis (an approximate Chebfun polynomial basis ) before being merged. In the end, we hope to map all imaging data processing to simple polynomial operations. Our massive data-sets required parallelizing some Chebfun functions on GPUs, allowing about 100 times faster polynomial evaluation and up to 12 times faster on CPUs when parallelizing the whole algorithm.
Highlights
The Institut de Recherche de Chimie Paris (IRCP6) and Centre de Recherche et de Restauration des Musées de France (C2RMF7) have set up a mixed research team, Physico Chimie des Matériaux Témoins de l’Histoire (PCMTH), in order to bring new solutions to the challenges facing the analysis, conservation and restoration of cultural heritage objects, and one of the team’s ambitious research projects concerns the latter’s digital images : how to best acquire, store, analyze and combine them
The C2RMF tools of the trade include optical and X-ray photography [16], and, thanks to the expertise brought by the team’s IRCP component in Electron Paramagnetic Resonance (EPR) who pioneered the application of EPR imaging (EPR-I) to exobiology [14], has decided to use EPR-I in order to image specific chemical species which X-rays and other traditional techniques are unable to target, like the different layers of carbon-related material found inside paintings
Data merging for the cultural heritage imaging based on Chebfun approach role Chebfun plays in our unifying approach to image data processing8, and how we managed to fit the key tomographic process of backprojection into the Chebfun paradigm ; we show that this approach provides a promising imaging data processing unification without having to pay a performance cost : it is a zero-cost abstraction9
Summary
The Institut de Recherche de Chimie Paris (IRCP6) and Centre de Recherche et de Restauration des Musées de France (C2RMF7) have set up a mixed research team, Physico Chimie des Matériaux Témoins de l’Histoire (PCMTH), in order to bring new solutions to the challenges facing the analysis, conservation and restoration of cultural heritage objects, and one of the team’s ambitious research projects concerns the latter’s digital images : how to best acquire, store, analyze and combine them. The C2RMF tools of the trade include optical and X-ray photography [16], and, thanks to the expertise brought by the team’s IRCP component in Electron Paramagnetic Resonance (EPR) who pioneered the application of EPR imaging (EPR-I) to exobiology [14], has decided to use EPR-I in order to image specific chemical species which X-rays and other traditional techniques are unable to target, like the different layers of carbon-related material found inside paintings This EPR-I information needs to be merged with the one gathered from other sources, be it X-rays or optical photography in order to provide a single object to which we may attach an interface for subsequent manipulations. We describe how we managed, by working both on the algorithmic and hardware aspect, to accelerate our Chebfun paradigm application : the execution time speed was scaled down by orders of magnitude compared to our original straightforward implementation on general purpose hardware
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
