Modeling the Past: Creating 3D Models from Archival Imagery
Written by Owen Murray, photographer at the Epigraphic Survey of the Oriental Institute, University of Chicago
Since the inception of photography, Egypt and its monuments have found themselves the subject of countless photographers, not to mention millions upon millions of tourists. The photogrammetric processes used to generate 3D datasets that have become a staple in modern Egyptological documentation require multiple, overlapping images, generally taken in a specific order. Although archival photographs rarely overlap enough for them to be used in this process, there are rare — and valuable — occasions when they do.
This article details the reconstruction of the East Wall of the Hall of Offerings (VIII) in Luxor Temple from the archival imagery of French Egyptologist Alexandre Moret, courtesy of the Collège de France. It covers how archival film negatives can be integrated with more recent digital imagery in order to produce rectified 3D models that virtually reconstruct monuments as they would have appeared in the past. It covers basic photogrammetric methodology, from image acquisition to post-processing and treatment of individual images using Adobe CS applications, through to combined image alignment, dense cloud, mesh and texture generation using Agisoft Metashape.
Alexandre Moret, 1868-1938 (Photo: Wikipedia).
Alexandre Moret was a French Egyptologist, member of the French Academy (1927) and Professor of Egyptology at the Collège de France (1923). He photographed the east wall of Luxor Temple over a period of 23 years (1914-1937), motivated in large part by the inaccuracies in Albert Gayet's publication Le temple de Louxor (1894). Moret planned to produce a new publication of Amenhotep III material in Luxor Temple to correct these mistakes, but unfortunately this was never realized. The manuscript of the first volume of his publication was sent to the French Institute in Cairo, but Moret's sudden death at the beginning of the WWII stopped the project — both manuscript and printing proofs were lost in progress.
What remains are Moret's photographs, housed in the Collège de France.
Moret took 43 B&W negatives of the Eastern Wall, fully documenting all 19 scenes and upper frieze material. Although photogrammetry in its current form wasn't a reality, he shot with enough overlap to cover the entire wall, except for the bottom 1.25m where there was/is no inscribed material.
Because each photograph was considered a composition unto itself, often focusing on just one or two scenes, lighting conditions between photographs and scenes very greatly. The physical state of many of the negatives also contributes to a discrepancy in quality.
Of the 43 B&W negatives, 27 were selected and level adjusted in Adobe PS, in order to balance them roughly, before then importing them into Agisoft Metashape alongside a set of 167 digital photographs taken in 2018.
Although Moret's photographs provide (almost) complete coverage of the entire wall, there was insufficient overlap between them to produce enough object geometry for a 3D model, thus the need for more recent digital photographs taken with the requisite overlap, and photogrammetric best practices in mind. The set of digital photographs were taken with a series of 6 coded targets and measured for scale-bar calibration in Metashape.
Combining the archival and digital photographs together not only allowed the proper alignment and rectification of the film photographs, but also provided sufficient information to generate a dense cloud and then mesh, onto which Moret's archival photographs were applied as the texturing source.
A series of 75 hand-picked points of interest were identified and used as targets to rectify and align the B&W negatives with the more recent digital photographs. The placement of these points was refined as best as possible to obtain an accurate match.
The bulk of the work in this model took place in alignment, slowly refining the number, and placement of targets. After refining the alignment, a dense cloud was built and then a mesh. Before texturing the mesh, all of the digital photographs were disabled so that only Moret’s 27 archival photographs, taken between 1914-1937, would be used when generating this texture.
After building the model, an orthomosaic was extracted from it, and the draw patch tool was used to select the best possible photograph/s for each scene on the wall.
The orthomosaic was then updated and exported as a high res jpeg at a 0.845 millimetre to pixel ratio. This mm/px ratio delivers a 1-10 scale @300dpi.
The exported orthomosaic was opened in Adobe PS and burned and dodged to further balance out areas where the discrepancy in lighting conditions of individual photographs was apparent between scenes. Following this, a curves level adjustment was applied to the entire image.
It’s impressive — Moret’s work allows for 1:10 @ 300dpi, which is pretty darn good, all things considered!
Click photo to enlarge (Click to zoom in, click and hold to move around)
To make the quality of this model even better, a further step would be to re-project the post-processed orthomosaic back onto the model, and then use it as the sole texture. Not to mention tweaking the very bottom left hand corner and adding in a few more targets so that the wall stops bending subtly!
The East Wall of the Hall of Offerings C.1930 offers a unique glimpse back in time, and a chance to encounter the wall as one might alongside Moret. It is one of many possible layers of interpretation and documentation that can add to our knowledge and understanding of the Amenhotep III material at Luxor Temple, not to mention any change in conditions over the past 100 years. It is thanks to the work of archivists like Alain & Emmanuelle Arnaudiès and institutions such as the Collège de France & The Epigraphic Survey that the work of a past generation of researchers and Egyptologists can find use in the present. Likewise it is thanks to those very same institutions and individuals that the work of the current generation will be made useful in the future — all through the proper use of keywords, captions and tags. Publish and archive your work, make it accessible, tag your images!
WHAT TO READ NEXT
Recording Djehutihotep. Digital epigraphy in a Middle Kingdom governor’s tomb at Dayr al-Barsha (Part 2)Written by Toon Sykora
After our preliminary study was completed, we set out to fully document the preserved decoration in the tomb of Djehutihotep. With originally more than 250 m² of painted surface and a high degree of detail in the decoration, this poses quite a challenge.
Some of our Egyptological work at Harvard centers around 3D modeling as a research and teaching tool, digital epigraphy, graphic design and book production. Our Giza Project at Harvard, now in its twentieth year...
Owen MurrayApril 17. 2020
To follow up with questions from the presentation that I wasn't able to get to:"Could this [process] be used for more monuments?" (Carolena Larson) Yes, I definitely believe that it can, but this depends on diving back into the archival record to see what's available and how it might be best applied. In some ways the thinking behind this is similar to a 'crowd-sourced' model approach.
Owen MurrayApril 17. 2020
"[Could one achieve the] same results without the new colour images?" (Mohammed Abdel Laziz)No, it wouldn't be possible -- and this has to do with how Moret photographed the wall in the 1930's -- there just isn't enough overlap between photographs, especially around the outer edges of the wall, to derive acceptable object geometry. Something needs to be used to provide that object geometry, whether it's a laserscan, or calibrated and scaled photogrammetric model, or any type of blend between the two. In this case I used the digital set of photos and photogrammetry, with long scale-bar measurements for proper calibration, to create the object geometry (model) necessary to make use of the archival images.
Owen MurrayApril 17. 2020
"[What was the] reference system used?” (Ibrahim Mustafa) In this case, although there is a local coordinate system in place for Luxor Temple, I didn't use X, Y, Z coordinates in 3 dimensional space in any reference system (local or otherwise) for the model. I used X, Y measurements in one plane (2 dimensions) in order to properly scale and calibrate the model. I wanted to demonstrate that one doesn't need traditional survey or laser scan datasets in order to produce this type of archival digital result. That said, the same results can be achieved using reference systems (local or otherwise) and indeed would be recommended if one were to attempt to reconstruct more than one wall, or anything type of object that has more complex object geometry than a flat wall.
Owen MurrayApril 17. 2020
"What is the next stage?" (Fred Botha) Well, the next step is to re-project the finished orthomosaic back onto the model to achieve a better level of quality, but, that said, I'm told that Moret also photographed all the other walls in the Hall of Offerings, so it may be possible to virtually reconstruct the rest of the room. This would be really interesting to do, but time-consuming to be sure!
Owen MurrayApril 17. 2020
"What could be learned between the old and new/are there any plans for analysis?" (Gai Jorayev, Beatrix Clark, Julia Puglisi)These are excellent questions, but unfortunately this is where I bow out as I don't have the requisite knowledge! My strengths lie in being able to assemble these sort of models and help from a technical perspective. That said, I happen to work for, and with, an amazing cast of Egyptologists, Artists, Architects, Archivists, Stone Masons and Conservators whom are more than capable of taking these results to the analytical level. And, in putting this material out to the larger community and making it available, anyone who in interested in doing these sort of comparisons can do so!
Yarko KobyleckyJuly 08. 2020
Thanking you in advance.