📄 Showing the "Simple" version — "Advanced" is not ready yet. Add it to favorites to help prioritize it. Ask for the full read

A photo taken through frosted glass always comes out blurry. As light passes through a sample, it scatters, and the camera sees only a smudge instead of a sharp image. Scientists have long known how to recover an image from such a smudge, but old methods are finicky: they either get stuck on a wrong solution or require too many exposures, bombarding the fragile sample.

The new approach works differently. Instead of guessing the whole picture from scratch, the program starts with a rough physical model, like a pencil sketch, and learns to correct only the discrepancies. It's like an artist who doesn't repaint the entire portrait but uses light strokes to fix inaccuracies. Moreover, the corrections apply not only to the image but also to the microscope's settings: the program refines in real time exactly how light traveled through the instrument. This combination of physics and learning makes the reconstruction robust and reduces the radiation dose.

The team tested the approach on various types of microscopes, from benchtop optical to electron microscopes, where a beam of electrons replaces light. With the electron beam, they could see connections between nerve cells in a brain slice — a detail that previous methods showed less clearly. Such a tool will be useful wherever you need many images and minimal damage to the sample, from inspecting microchips to studying living cells.

Light that used to be a nuisance now helps bring things into focus. The next step is to teach microscopes to use this hint as easily as a photographer turns the lens ring.