Image system that spots fake Post-Impressionists
Dutch researchers have developed an analytical computer application which, it’s claimed, enables the better interpretive recognition of the content of images and data sets.
Called 'Reading Images for the Cultural Heritage' (RICH), the application forms part of a larger IT project called 'Continuous Access to the Cultural Heritage' (CATCH), initiated by the Netherlands Scientific Organisation (NWO) to foster use of IT in the country’s heritage sector. RICH is the culmination of a four-year long R&D programme.
“The technique is designed to visualise high-dimensional data - data that comprises many variables,” explains Laurens van der Maaten, who led the development team. “The data can be weather data, image data, or any data you can think of.”
RICH analyses the given data, and then computes pairwise distances between all measurements contained within it; the resulting pairwise distance table is comparable to that of a standard distance table between cities, for instance. RICH then constructs a data ‘map’, based on distances from the pairwise distance table.
The prototype image capture system consists of two arms - one to hold illumination, and one to hold a camera - and a software-operated turn-table that is precise up to 0.1º, its designers say. RICH’s software takes hundreds of photographs of an object in a two-minute sequence to compile the images on which its pairwise analysis is based.
“We have developed computer vision algorithms for the analysis of textures, shapes, and stamps,” van der Maaten reports. “[These are] of interest in the classification of flint, ceramics, stamps, and also paintings. We were able to discriminate between real and fake Van Gogh paintings by applying our texture analysis technique onto a dataset of high-res digital reproductions of paintings.”