Back To Schedule
Thursday, May 29 • 3:00pm - 3:30pm
(Research and Technical Studies Session) Accurate Measurement and the Quantification of Surface and Material Property Change Using New RTI and AR Techniques

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

This presentation examines new open source software that will dramatically improve the accuracy of the results generated by Reflectance Transformation Imaging (RTI) and Algorithmic Rendering (AR).

RTI and AR use the same set of empirically captured photos. This photo set is acquired with a fixed camera and subject. Each photo is taken with the illumination source in a different position. The positions evenly sample incoming light directions over the surface of an imaginary hemisphere over the subject.

Cultural Heritage Imaging (CHI) and a team at Simon Frasier University (SFU) will demonstrate new ways of using RTI data. The new RTI algorithm produces colors, self-shadowing, and specular highlights that exactly match photographic ground truth. The team removes one of the photos from the RTI set and synthetically generates a new photo from the same light position. The SFU team has used many different subjects to convincingly demonstrate their synthetically built photos exactly match the attributes of the removed photo. The new algorithm can generate accurate shadows and highlights associated with any incoming light direction.

The SFU algorithm generates highly accurate surface shape information. In RTIs, surface shape information is recorded as a field of surface normals. This field represents the spatial orientations of the subject’s surface at the locations corresponding to the photo’s pixels, which represent it. This orientation data determines the direction of light reflectance off the surface of the imaging subject, produced by an interactively pointed virtual light.

The SFU research algorithm uses a radial basis function to separate shadows and highlights and weights them according to their intensity. The algorithm also creates a third set of pixels, which have neither shadows nor highlights. The normal field is calculated per pixel using the pixel samples from this third set. The calculation of normal directions can be significantly misdirected by the presence of shadowed and highlighted pixel samples. Normal direction calculations using pixel samples without shadows and highlights produce highly accurate normal fields. These normal fields accurately represent the topology of the imaging subject's surface. These normal fields can also be integrated to create full 3-D geometry representing the subject's surface.

When this geometry is measured, it will yield accurate results. Once the subject is represented in a measurable form, subsequent RTI data sets can be transformed into measurable 3-D representations, which enable the accurate quantification of surface shape change. Measurement of surface color and material characteristics, such as shininess, can also be quantified to track their changes over time.

These new tools are being integrated into existing CHI open source software for building and viewing RTI's. CHI collaborators at Princeton University are building the photometric stereo, 3-D geometry building, open source software into the AR toolkit. An overview of new features to RTI and AR tools will also be presented.

These tools have the potential to dramatically improve the quantification of change of humanity's legacy under the stewardship of the conservation community.

avatar for Mark Mudge

Mark Mudge

President, Cultural Heritage Imaging
Mark Mudge is President and co-founder of Cultural Heritage Imaging (CHI) and the current Chairman of the Board of Directors. Since decades Mark has worked in 3D information capture environments and digital photography. Together with Tom Malzbender he is a co-inventor of the Highlight... Read More →
avatar for Carla Schroer

Carla Schroer

Director, Cultural Heritage Imaging
Carla Schroer is co-founder and director of Cultural Heritage Imaging (CHI) a non-profit corporation that develops and implements imaging technologies for cultural, historic and artistic heritage and scientific research. Carla leads the training programs at CHI along with working... Read More →

avatar for Mark S. Drew

Mark S. Drew

Professor of Computing Science, Simon Fraser University
Mark S. Drew is a Professor in the School of Computing Science at Simon Fraser University in Vancouver, Canada. His background education is in Engineering Science, Mathematics, and Physics. His interests lie in the fields of image processing, color, computer vision, computer... Read More →

Mingjing Zhang

Graduate Student, Simon Fraser University

Thursday May 29, 2014 3:00pm - 3:30pm PDT
Seacliff C-D