Optimizing compositional images of daguerreotype photographs using post processing methods

Abstract

Microfocused X-ray fluorescence imaging was used to examine two 19th century daguerreotypes. The distribution of Hg-bearing nanoparticles beneath the Au gilding layer gives rise to contrast in the photographs. The sum image for Au + Hg M line X-rays is therefore a compositional image that mimics daguerreotype photograph contrast. Because the thickness of the Au and Hg on the surface represents a small fraction of the X-ray activation depth, the resulting sum images are not of high quality and are dominated by Poisson noise. This is true even for long collection times up to multiple days in length. Achieving superior contrast resolution by increasing the duration of data collection further is impractical. In this study, a new image processing technique based upon the Haar-Fisz algorithm and wavelet theory was used to improve the image quality of compositional maps. The digitally processed images show a reduction in noise without a loss of spatial resolution over the length-scale of photographic features. The Haar-Fisz denoising algorithm decreased the contribution of the Poisson noise in compositional images. Multiple resolution analysis improved image quality further. Features within the portraits are uniformly more recognizable in the final processed images relative to the raw X-ray images. Intensity line profiles that traverse midtone, highlight and shadow regions of the daguerreotype reveal that spatial resolution is not degraded by the image processing routines. Improvement in image quality is quantified by comparing the relative variance of the raw and Haar-Fisz processed imagery. The use of a Haar-Fisz denoising transformation, coupled with multiple resolution analysis was found to improve the quality of low count X-ray images without impacting the spatial resolution at the scale of photograph features. The process can be implemented in freely available open-source software with a minimum of programming effort. Such digital post processing routines offset the need for longer acquisition times to achieve improved X-ray fluorescence image quality. Finally, because the Au + Hg map is insensitive to surface imperfections and tarnishing via atmospheric adsorption of sulfur, digitally processed images may be used to reconstruct photograph features in heavily disfigured daguerreotypes.