Sen—Last week, the team on the New Horizons space probe released amazing photos of Pluto: For the first time, we could see surface features on the tiny world!
I’ll admit I was surprised by the announcement. New Horizons was still a long way from Pluto, over 100 million km. At that distance, in the New Horizons camera the planet should be only a bit over 4 pixels across. That’s pretty small! But in the new images, Pluto is about 10 pixels across. How is that possible?
The answer involves statistics, digital camera engineering, optical phenomena, and even the Hubble Space Telescope!
The pictures of Pluto are taken by the Long-Range Reconnaissance Imager, or LORRI. Like all digital cameras, it converts light that hits its detector into electrons. These electrons accumulate in each pixel of the array like rainwater collecting in buckets. Over time, a two-dimensional picture is built up of the target. Where the target is brighter, there are more electrons in that pixel.
The electrons are then counted by the detector and converted into a table, a spreadsheet if you will, recording the number of electrons in each pixel. When your computer reads that table, it converts the numbers to brightness, and you see a picture.
Each pixel sees a small area of the sky. In the case of LORRI, each pixel sees an area about an arcsecond across. That’s pretty tiny; there are 3,600 arcseconds in a degree, and for comparison the Moon in the sky is about 1,800 arcseconds.
Stars are so far away that they are far smaller than one arcsecond in size. You’d therefore expect that a star would only fill one LORRI pixel, and look square in the picture. But in reality things are more complicated. When light enters a telescope and is focused onto the detector it spreads out a bit due to diffraction. When all is said and done, the star actually spreads out over several pixels, looking like a chunky circle.
This is true even for bigger targets, like Pluto. Even though technically Pluto was only 4 arcseconds across, the light was spread out a bit, making it look bigger. Yet we see features far smaller than this on its surface! It turns out you can see features that are actually smaller than single pixel if you’re clever and plan your observations well.
Think of it this way. Imagine you have two stars very close to each other in the sky. If you use a camera with big pixels, a single pixel might cover them both. When you look at the picture, they smear into a single blob.
But if you have extremely teeny pixels, then you can cleanly separate them in the final image. Astronomers call this ability to separate out close objects resolution.
Now imagine you have a camera with medium size pixels and you look at those two stars. They’d look slightly elongated in your picture, smeared together but not circular. There’s a trick you can use to see them better: Scan the camera across them, taking a picture every time you move the camera some small fraction of a pixel. So: Move the camera say ½ the width of a pixel, take a picture; move it again, take another. Lather, rinse, repeat.
If you then take these individual pictures and combine them carefully, you can create a bigger picture with higher resolution then you can in any single individual picture! This is called superresolution. It works, but is a bit limited. You can’t really use it to infinitely zoom something, but it can double the resolution of your camera.
Surface features can be seen to change as Pluto rotates in an animation made of several sharpened images. Image credit: NASA/JHUAPL/SWRI/Tod Lauer and Constance Tsang
That’s why Pluto looks twice as big in the New Horizons pictures than I expected. The team knew they could use superresolution to get better pictures even when the probe was still far away from Pluto, and took lots of images of the tiny world with sub-pixel movements of the camera.
But then they used another trick, called deconvolution. Think of it this way: We know a star should look like a point of light in the camera, smaller than a single pixel. Due to the nature of light, it spreads out. But if we know exactly how the light gets spread out, we can apply some math to “suck” that light back into the pixel, improving the resolution even more!
This technique is actually pretty complex, and has many pitfalls for the unwary user, but careful application can be effective. To perform deconvolution, you need very good images of stars, so you know just how a point source spreads out. Again, the New Horizons team (headed by astronomer Alan Stern) anticipated this need, and took lots of images of star clusters to make sure they had a good map of how light spreads out across the LORRI detector.
From blurry raw images of Pluto to seeing surface features. Image credit: NASA/Johns Hopkins University Applied Physics Laboratory/Southwest Research Institute / Tod Lauer
The software to perform the deconvolution was written by astronomer and programmer Tod Lauer, based on known mathematical techniques (for the imaging geeks out there: the Lucy-Richardson method). And he was a good choice: He wrote some of the original software that did exactly this when Hubble Space Telescope was launched in 1990!
The telescope had a flaw in the mirror that made it out of focus, and Lauer wrote the software to help sharpen those images. In fact, I was using Hubble for my PhD research when we got those first images back, and I used Lauer’s software (which I modified for my own needs) to deconvolve the images. They weren’t perfect, but they were pretty good.
And that’s how New Horizons got such sharp images of Pluto, even so long before the actual close encounter. As Lauer pointed out to me, LORRI uses a telescope that only has 21 cm mirror. That’s the same size as my own backyard ‘scope, and Lauer has a 25 cm ‘scope of his own! But sometimes, like real estate, astronomy is all about location.
And property values are about to go up. As New Horizons approaches Pluto, and the little iceball appears bigger and bigger, the resolution will improve as well.
What we will see? No one really knows… but thanks to Lauer and the New Horizons team, what we do see will look amazing.