NASA's Mars Imaging Algorithm Is Now Decoding Ancient Rock Art

A color-enhancement algorithm built at NASA’s Jet Propulsion Laboratory to sharpen satellite and planetary imagery has found a second life in archaeology. The method, called decorrelation stretch, was devised at JPL in 1978 when Jim Soha proposed applying the Karhunen–Loève Transform — a statistical technique for concentrating a signal’s information into fewer components — to digital color imagery. Rather than merely boosting contrast, it remaps the original colors onto an expanded range, pulling out detail the human eye would otherwise miss. Ronald Alley refined the process in 1996 for speed and accuracy, and his write-up became the crucial documentation that carried the technique beyond NASA.

Archaeologist Jon Harman, a retired researcher with a Berkeley math PhD and a background in medical imaging, encountered the technique at a 2005 rock-art conference after seeing false-color images of Mars. He tracked down Alley’s paper, coded it into a plug-in he named Dstretch, and added custom color spaces tuned for faded pigments. An early test on photos from Baja California surfaced a yellow figure that had been effectively invisible, convincing him of its value.

The tool has since enhanced hundreds of images and produced concrete discoveries: bat and pig figures at Egypt’s Beni Hassan cemetery, a Crow warrior’s taunting horse-and-rider pictograph in Alberta’s Writing-on-Stone Provincial Park, and roughly 15 previously undocumented figures plus new detail in 28 others at Norway’s Årsand 1 site. The story is a clean example of dual-use engineering — a remote-sensing algorithm crossing disciplines with little more than a published paper and someone willing to reimplement it.