Old Photographs for New Geometries: The Recent Evolution of Photogrammetry

Back to overview

A look back

Over the years, 3D photogrammetry has proven to be the most reliable and democratizing technology for heritage virtualization. Today we are capable of capturing with sub-millimeter precision everything from very small objects that are less than one centimeter long to large structures in multi-faceted projects with thousands of input images. However, the road traveled has been intense: originally photogrammetry was not as effective as it is now.

I remember doing my first photogrammetric tests between 2011 and 2012. The results left a lot to be desired, but for someone who was approaching the virtual world for the first time, they already represented a tremendous qualitative advance. Fast forward a few years and technical evolution was leading to improved software and algorithms, especially from Agisoft Photoscan—the predecessor of Agisoft Metashape. Photogrammetric models achieved increasingly higher levels of detail, so much so that they came to compete with the exceptionally detailed laser scanners.

Skipping forward in time to the present day, what we lived in the years 2015–2017 now seems outdated to us. The most recent changes and improvements have really further developed this technology. It seems possible that we are already reaching a consolidation plateau given that now developers are focusing more on reducing processing times and refining small details.

Revisiting old photosets

In any case, the theme that makes me write these lines and that takes me down the path of virtual memories is of special importance. I always thought that photogrammetry would achieve the quality that we enjoy today and in that thought, back in 2015 I started to save and organize raw image data from my first professional projects. All the raw images have since been stored on various hard drives in the hope that they may be used again at some point. Good decisions are valued and best enjoyed in the medium and long term. This was one of them because that time has come.

This is one of the many examples of virtualized pieces from “those distant and archaic times”. Now, five years later, I have decided to reprocess, comparing the results of then with those of now. This is the Bicha de Balazote Iberian protohistoric statue from the National Archaeological Museum of Spain, photographed on April 20, 2015.

bicha of balazote comparison 2015 2020

The differences are obvious. The geometric details are much more defined, the colors are more realistic and the finishes more precise. And all this without having taken a single new photograph. With the same data but the latest technology, spectacular results can be achieved. Even the optimization and lighting techniques have been greatly improved, allowing us to show all the detail with a small percentage of the weight, and eliminating uneven shadows and highlights.

It seems incredible that the two models—2015 on the top and 2020 on the bottom—have been created with the same set of photographs. Without a doubt, this should make us reflect on what the future holds for us. Perhaps the changes will be less noticeable from now on, but I’m sure they will come. In the meantime, I will continue to keep terabytes and terabytes of information from all my photogrammetric projects to get new joys like this in the not-too-distant future.


About the author

Néstor F. Marqués

Spanish archaeologist devoted to virtual Heritage and cultural dissemination. Passionate about ancient Rome and its culture. More info: nestormarques.com

1 Comment

  • Dale Utt says:

    This is such a useful exercise. I’ve found that it can become rather easy to miss all the incremental progress the software (and your own technique) has made if you’re doing photogrammetry every day or every week. Being able to take a step back and look at my first models versus now, the difference is pretty stunning. You have also made me feel validated in obsessively saving all my old data sets!

Leave a Reply

Your email address will not be published. Required fields are marked *

Related articles