Learning computer promises sharpest images

A computer system that is able to learn like a human brain could be used in future to push past the limits of optical telescopes and view deep-sky objects in greater detail.

In 1995, the majestic spiral galaxy NGC 4414 was imaged by the Hubble Space Telescope as part of the HST Key Project on the Extragalactic Distance Scale. An international team of astronomers, led by Dr. Wendy Freedman of the Observatories of the Carnegie Institution of Washington, observed this galaxy on 13 different occasions over the course of two months. Images were obtained with Hubble's Wide Field Planetary Camera 2 (WFPC2) through three different colour filters. Based on their discovery and careful brightness measurements of variable stars in NGC 4414, the Key Project astronomers were able to make an accurate determination of the distance to the galaxy.

Spiral galaxy NGC 4414, imaged by the Hubble Space Telescope in 1995. Will this new technique enable astronomers to go back over old data and see the objects more clearly? Credit: Hubble Heritage Team (AURA/STScI/NASA/ESA)


Every astronomer is limited by the size of their telescope’s lens or mirror.

The bigger the mirror, the more light it can gather and the better it can see faint objects in the sky.

But what if a computer programme was able to take an image captured by a telescope and calculate how the target object would look if seen in greater detail?

A group of Swiss scientists are hoping their new programme can do just that.

Using a system called ‘neural nets’, which enables computers to learn in the same way as a human brain, the team began by showing the computer a blurred and a sharpened image of the same galaxy.

It was then able to use this knowledge to clarify a blurred galaxy image by itself.

Not only this, the computer was also able to recognise and resolve features that the telescope could not, like dust lanes, bars and star-forming regions.

This technique, once refined, could enable astronomers to see farther into space than the light-gathering capabilities of the biggest telescopes allow.

“We can start by going back to sky surveys made with telescopes over many years, see more detail than ever before, and for example learn more about the structure of galaxies,” says Professor Kevin Schawinski of ETH Zürich, who led the study.


“There is no reason why we can’t then apply this technique to the deepest images from Hubble, and the coming James Webb Space Telescope, to learn more about the earliest structures in the Universe.”