Enhancement of hidden patterns in paintings using statistical analysis

In this paper, we discuss the application of different Blind Source Separation algorithms for enhancing hidden patterns in paintings and retrieving the information there contained. A series of images was taken at different wavelengths of a painting of an unknown artist, which evidenced the presence of hidden text. The analysis of the single images did not allow retrieving the information contained in the hidden text. The application of Blind Source Separation algorithms on the other hand allowed the enhancement of the hidden pattern; in particular, the orthogonalization algorithm proved to be fast and highly efficient. Good results were also obtained applying the same algorithm on a subset of spectral images (two out of four available), demonstrating the possibility of further reducing the computational load of the algorithm while retaining a good readability of the retrieved text.