A faded painting may look almost indistinguishable from the rock beneath it while a photograph still records small color differences. Digital enhancement can amplify those differences, making an outline or patch of pigment easier to inspect. The vivid result is useful evidence about the image, but it is not a restored view of what an ancient artist originally saw.
One widely used approach is decorrelation stretch, available to researchers through the DStretch program. A September 8 technology feature traces its connection to remote-sensing methods developed at the Jet Propulsion Laboratory. Jon Harman adapted the approach for archaeological imagery in 2005.
The feature includes paintings at Angkor Wat identified during work in 2010–2012. Those finds are earlier examples of the technique’s use, not discoveries made this September. Their relevance is methodological: apparently bare or badly faded surfaces can contain visual information that ordinary viewing fails to separate.
Keep exploring
Continue with Curiosmos
What the software changes
A digital color photograph stores information in channels. On a reddish rock face, the background and a fading red pigment may vary together so strongly that the distinction is hard to see. Decorrelation stretch transforms those relationships and expands differences that were compressed into a narrow range.
Harman’s description of the algorithm explains that the process produces false color. A conspicuous blue, green or purple area in the output need not have been painted that color. The purpose is to distinguish features, rather than reproduce the original palette.
The method generally works better for pictographs, made with pigment, than for petroglyphs cut into rock. An engraving may depend more on relief and lighting than on differences in hue. The right imaging approach therefore depends on what survives at the site.
Why two enhancements can look different
The selected color space and the colors included in an image affect the transformation. A result is not a fixed revelation independent of the settings. Researchers need to compare it with the unprocessed photograph and understand which choices made a feature more visible.
A 2026 study of localised decorrelation stretch addresses another limitation: a whole image can contain such a wide distribution of colors that subtle differences in a smaller region remain poorly separated. Benedict Dyson, Andrea Jalandoni and Paul Taçon developed an approach that processes overlapping windows before merging them.
The team’s method documentation reports improved visibility of fine details in densely painted areas. The underlying point is practical: processing a large, varied rock-art panel as one unit is not always the most informative way to inspect each part of it.
Visibility is not a date or an identification
Enhancement cannot independently establish who made a mark, what it represents or how old it is. Those questions require archaeological context and, where suitable material survives, dating evidence connected to the painting. A clearer outline does not supply a missing chronology.
Nor can the process retrieve a motif that left no recorded signal in the photograph. Its value lies in making existing differences easier to examine. Interpreting a newly visible shape still requires checking the original image, the physical surface and alternative explanations for the discoloration.
A useful published comparison therefore keeps the original beside the processed version, identifies the settings and describes the specific feature under discussion. If only the spectacular false-color image is shown, readers lose the means to judge how much of the result comes from a subtle surviving trace and how much from the way that trace has been displayed.
