A photo-editing method NASA engineers first built to sharpen satellite images of Mars and Earth has quietly become one of archaeology’s most useful tools for recovering pictures that faded almost to nothing. Known as decorrelation stretch, the process exaggerates subtle color differences that the human eye can no longer separate, pulling detail out of rock paintings, aerial photographs, and other historic imagery that otherwise looks like blank stone or worn pigment. The technique traces back to a 1978 image-processing paper written at NASA’s Jet Propulsion Laboratory, but it did not reach archaeologists until a retired medical-imaging specialist adapted it for his own rock art photographs in the mid-2000s. Since then, the tool has uncovered hidden paintings, long-forgotten graffiti and buried building foundations on several continents.
How Decorrelation Stretch Separates Colors the Eye Can’t
Decorrelation stretch does not simply raise contrast the way a basic photo filter would. It maps an image’s original colors onto a wider, mathematically remapped range of colors, a process that involves diagonalizing color matrices using a statistical method called the Karhunen–Loève Transform. JPL researcher Jim Soha co-wrote the 1978 paper that first proposed applying the transform to digital imagery for color enhancement, and his colleague Ronald Alley later refined it, replacing an error-prone multi-step process with a single matrix multiplication that made the technique faster and more precise. Alley wrote up the method in a 1996 paper while helping develop the Advanced Spaceborne Thermal Emission and Reflection Radiometer, an instrument built by Japan that launched aboard NASA’s Terra satellite in 1999 and is still operating today. Geologists at JPL had already been using an earlier version of the technique to map lava flows in Hawaii before the instrument ever reached orbit.
A Mars Rover Photo That Convinced a Rock Art Hobbyist
The technique might have stayed a niche tool for geologists and volcanologists if not for Jon Harman, a rock art enthusiast in Pacifica, California, who also happened to work in medical imaging. Around 2005, someone at a rock art conference showed him NASA images of the Martian surface, presented with and without decorrelation stretch applied. Struck by how much extra detail the processed version revealed, Harman searched for the underlying method, found Alley’s paper, and recognized that his own imaging background meant he could build it himself. He wrote a plug-in called Dstretch for ImageJ, an open-source image-analysis program originally developed by the National Institutes of Health. His first real test came on a photograph from the Cave of San Borjitas in Baja California, Mexico, where a previously invisible yellow human figure emerged in the middle of the frame once the algorithm ran. That result, he has said, is what convinced him the tool was worth pursuing further.
Faded Paintings High Inside Angkor Wat
One of Dstretch’s best-known finds sits near the top of the central tower at Angkor Wat, the sprawling Cambodian temple complex that draws thousands of visitors a day. Paintings there depict horseback riders and a traditional Cambodian musical ensemble called a pinpeat, faded to the point that passersby walk beneath them without noticing anything is painted on the stone at all. Between 2010 and 2012, archaeologist Noel Hidalgo Tan used Dstretch to identify these images along with roughly 200 other paintings scattered throughout the temple complex, work that turned an overlooked corner of one of the world’s most-photographed monuments into a documented gallery of centuries-old artwork.
From Norway’s Rock Carvings to Egyptian Tombs
Similar rediscoveries have followed at rock art sites well beyond Southeast Asia. At the Årsand 1 site in western Norway, pictographs including a sun symbol and stylized human figures were first documented in 1940, but applying Dstretch in 2008 and 2012 revealed about 15 previously unrecorded figures and new detail in 28 others already known. In the ancient Egyptian cemetery of Beni Hassan, archaeologists used the tool to identify painted bats and pigs, animals that rarely appear in Egyptian art and were easy to miss in weathered tomb interiors. At Writing-on-Stone Provincial Park in Alberta, Canada, Dstretch brought out what researchers interpret as an early form of tagging: a pictograph of a horse and rider believed to be a Crow warrior’s calling card, left to taunt rival Blackfoot groups.
Buried Ruins and Tattoos on Mummified Remains
Dstretch’s usefulness has extended past painted surfaces into other kinds of faint or buried evidence. At the ancient settlement site of Vlochos in Thessaly, Greece, occupied between roughly 500 B.C.E. and 800 C.E., researchers applied the technique to aerial imagery and turned up potential archaeological features that ground-penetrating radar, electrical resistance surveys and elevation models had missed, including structures obscured by metallic contamination in the soil. In 2023, researchers from the University of Tübingen and the Tennessee Division of Archaeology published a formal protocol for using Dstretch to sharpen faded tattoo imagery preserved on mummified human remains, extending a technique built for reading mineral signatures on Mars into the study of ancient human skin.
A One-Person Operation, Two Decades Later
Harman still runs the processing side of Dstretch largely by himself, fulfilling roughly 200 image-enhancement requests a year from researchers around the world for a $50 fee each. Around 2010 he built a simplified version as a smartphone app, priced at $20 and downloaded thousands of times, that mimics decorrelation stretch without the heavier computation the full desktop process requires. He has said he expected rock art researchers to embrace the tool, given how well it worked on his own image collection, but that the range of unrelated fields that adopted it, from Egyptology to forensic tattoo analysis, still surprises him.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
