Spectral RTI

Todd R. Hanneken, Saint Mary’s University

Brill Textual History of the Bible Volume 3

Pre-publisher copy last revised May 2017

Spectral Reflectance Transformation Imaging (Spectral RTI) combines the advantages of Spectral Imaging (→ Spectral Imaging) with the advantages of Reflectance Transformation Imaging (→ RTI) into a single consistent data set. For each pixel both color and texture properties are known and can be processed and visualized together. The major advantages gathered from spectral imaging are improved spatial resolution, color resolution, color range beyond the visible spectrum, processing enhancements, and objectively accurate color. The major advantages gathered from RTI are interactivity and enhanced visualization of texture. The combination is greater than the sum of its parts in as much as color and texture information are complementary. For example, even a single letter in an illegible manuscript may have some trace of writing from remaining ink and another trace in the outline of the corrosion into the surface of the parchment where ink had once been. A human user can visualize the information together in a single view for a more complete and natural experience of the manuscript. The consistent combined data also has potential to improve machine learning applied to manuscripts. The cost in time and equipment of Spectral RTI is less than the sum of each. As of 2017, spectral imaging systems can vary from a few thousand to over a hundred thousand dollars, with options for adding RTI of comparable sophistication adding approximately ten percent across the range of cost. There is no cost for the processing software. The cost in time is less than the sum of each because mounting, focus, metadata entry, etc., is done only once for the complete data set. Capture of a single page may take a minute to more than twenty minutes depending on the equipment, resolution, and thoroughness selected.

Spectral RTI aids textual criticism of biblical literature in its ability to recover text and other information from erased, damaged, or otherwise illegible manuscripts. Spectral RTI has advantages over first-hand experience in that color and texture can be enhanced through digital processing to greatly augment natural human perception. Spectral RTI also has advantages over conventional digitization in its interactivity with respect to texture and glimmer (specularity) and objective accuracy of color. These advantages aid the recovery of text as well as conservation and the study of the creation and use of the manuscript as an artifact of scribal culture in its own right. Traces of ink that may be indistinct from their surroundings to the human eye become easily distinct if they have different spectral signatures, as do different materials of similar color. The outline of letters can be read from the corrosion of slightly acidic ink into the surface of the parchment. Scoring lines used by scribes can direct the reader to where text should be. Warp and curl can be relevant to identifying a letter as well as tracking conservation. The ability to distinguish rise from recess, accretion from wear aids the identification of materials and purpose. In addition to the main text, Spectral RTI can help distinguish different inks from different stages of scribal addition, marginalia, dry-point notation, evidence of liturgical use, colophons and notes of ownership erased as the manuscript changed hands. All of this information about materials, production, and use over time can also be relevant to identifying forgery.

The underlying principle of the integration of Spectral and RTI is the distinction between chrominance (color) and luminance (lightness). Spectral imaging, though variously defined, fundamentally entails attention to wavelengths on the electromagnetic spectrum within or near the range visible to humans. Simply put, spectral capture and processing are concerned with color beyond the natural abilities of the human eye or conventional digitization. Texture imaging such as RTI entails attention to changes in luminosity at specific light positions. RTI is concerned with highlights, shadows, and glimmer as light moves over a texture, which are properties of brightness. Spectral and RTI technologies are compatible because of their discrete concerns. The combination of chrominance information from spectral imaging with luminance information from RTI capture is easily achieved using color spaces which orient their axes for luminance and color, such as YCbCr. An example of the YCbCr color space in action is the addition of color to previously monochrome broadcast television. A black and white television could continue to utilize the luminance (Y) sub-channel, but a color television could use the same luminance sub-channel and “colorize” it with the addition of two chrominance sub-channels (Cb for degree of blue chrominance and Cr for degree of red chrominance). The expression of light as a combination of luminance and two axes of chrominance is easily converted to other color spaces, such as RGB which orients its three axes as brightness in the red, green, and blue ranges respectively. In practice, this means that any image produced by spectral imaging with diffuse light can be recreated for any given light position by replacing the luminance channel in the diffuse light image with luminance from a particular light position. Because only luminance is used for any given light position the hemisphere captures for RTI need only one channel. This allows the higher-quality panchromatic sensors of spectral imaging systems to capture the needed data without interference filters or other effort to distinguish chrominance at each light position.

Spectral Reflectance Transformation Imaging begins with capture. It is effectively impossible to apply Spectral RTI processing after the fact to images that were not captured with a complete set in one session. It is essential that the object and camera remain immobile throughout the sequence so that each pixel represents the same point on the object throughout the data set (registration). Spectral imaging and RTI each require this consistency individually; for Spectral RTI the consistency must be maintained for both. The options for capture equipment are the same for the combination as they are for each individually, with the only additional concern that the apparatus for one should not interfere with the other. As is typical for spectral imaging, the lighting should be diffuse (coming from two or more panels with diffusers). As is typical for RTI, the light positions should be at a consistent distance from the object, at least four times the radius of the object being imaged. This can be achieved with a dome with lights at fifty or so positions, an arc that moves banks of lights to different segments of a virtual dome, or a handheld flash positioned at each point in the virtual dome individually. MegaVision Inc. produces an arc with an eight-foot (2.4 meter) diameter that captures up to sixteen (typically eight) light positions in each of seven segments for a total of typically fifty-six light positions evenly distributed around a virtual dome. The arc is more portable and flexible in the workspace than a full dome. A hand-held flash would be less expensive but require more time to position the flash and calculate the light position from a reflective hemisphere in the image frame.

Once the data is captured the image processing can be completed using open-source software. As of 2017, the Spectral RTI Toolkit is being actively developed and published on GitHub with support from a grant from the U.S. National Endowment for the Humanities. The Toolkit works with ImageJ, an image processing suite initiated by the National Institutes of Health. Open-source plugins provide colorspace conversion and Principal Component Analysis for processing spectral data. Commercial alternatives, such as ENVI, can contribute additional custom color processes. The RTIBuilder software distributed by Cultural Heritage Imaging can be used to calculate light positions (an .lp file) from highlights on a shiny sphere in the image frame (if not already known from an arc or dome). RTI files can be generated using the open-source Hemispherical Harmonics method or the free but not open-source Polynomial Texture Map method. RTI files can be published for viewing in any web browser (no plugins or downloads required) with the open-source webGLRTIMaker tool maintained by Gianpaolo Palma. The Spectral RTI Toolkit oversees all these components from a graphical user interface with no editing of text files or command line arguments required. Once created, the Spectral RTI images can be viewed in the same ways as conventional RTI, namley through standalone viewers or WebRTI in a web browser. As of 2017, textures created using RTI have not yet been incorporated into major 3D engines such as Unreal and Unity.

Creative Commons licensed examples of Spectral RTI are available online from the Jubilees Palimpsest Project using International Image Interoperability Framework (IIIF) Presentation manifests (http://jubilees.stmarytx.edu/iiifp/). The 2013–2014 phase established a proof of concept using test objects of varying degrees of texture, from a palimpsest to a terracotta figurine. The 2016–2019 phase brought the technology to the Biblioteca Ambrosiana for imaging of palimpsests. As of 2017, Spectral RTI images are available for the entirety of C73 inf (including Jubilees, the Testament of Moses, and an Arian Commentary on Luke) and sample pages from Origen’s Hexapla, Wulfila’s fourth-century translation of the letters of Paul into Gothic, and an unidentified Greek commentary on Luke.

Bibliography

Todd R. Hanneken, “Integrating Spectral and Reflectance Transformation Imaging for the Digitization of Manuscripts & Other Cultural Artifacts,” NEH Office of Digital Humanities White Papers (2014), https://securegrants.neh.gov/PublicQuery/main.aspx?f=1&gn=HD-51709-13.

Todd R. Hanneken, “New Technology for Imaging Unreadable Manuscripts and Other Artifacts: Integrated Spectral Reflectance Transformation Imaging (Spectral RTI).” In Ancient Worlds in a Digital Culture. Edited by Claire Clivaz, Paul Dilley, and David Hamidović. Digital Biblical Studies 1. Leiden: Brill (2016), 180–195.

Todd R. Hanneken, “Digital Archaeology’s New Frontiers,” Biblical Archaeology Review. March/April 2017. Pages 28–29, 65–66.

Todd R. Hanneken, “SpectralRTI_Toolkit,” GitHub, https://github.com/thanneken/SpectralRTI_Toolkit.

Todd R. Hanneken, “Spectral RTI Technology,” The Jubilees Palimpsest Project, http://jubilees.stmarytx.edu/#spectralrti.

See also the Bibliography at the separate entries for (→ Spectral Imaging) and Reflectance Transformation Imaging (→ RTI) individually.

Keywords

Spectral imaging; hyperspectral; multispectral; Reflectance Transformation Imaging (RTI); Polynomial Texture Mapping (PTM);

UCLA Rouse Ms. 32 Palimpsest, Accurate Color, diffuse light UCLA Rouse Ms. 32 Palimpsest, Accurate Color, raking light position UCLA Rouse Ms. 32 Palimpsest, PCA Pseudocolor, raking Figure 1: UCLA Rouse Ms. 32 imaged by the Jubilees Palimpsest Project 2013. Shown with variants for Accurate Color diffuse lighting (1a), Accurate Color raking light (1b), and PCA Pseudocolor processing raking light (1c). USC Antiphonary, high resolution Figure 2: USC Antiphonary imaged by the Jubilees Palimpsest Project 2013. At high spatial resolution with raking illumination the grain of the parchment becomes visible. Latin Moses, scribal scoring Figure 3: Biblioteca Ambrosiana C73 inf (The Testament of Moses) imaged by the Jubilees Palimpsest Project 2017. With raking light the scribal scoring lines of both the original fifth-century manuscript and the eighth-century palimpsest are evident. PERMISSION REQUIRED FROM THE BIBLIOTECA AMBROSIANA FOR COMMERCIAL USE