New Zoom-and-Enhance Technique Brings Mars Into Focus
"Zoom and enhance" is a cliche of TV cop dramas. But two astronomers have just discovered a way to do it in space, squeezing out an insane amount of resolution from still, satellite images of Mars.
"With just eight images, each with a reasonable quality, we can get up to five times more resolution."
Peter Muller and Yu Tao of University College London developed a computer program that combines aerial satellite photos of Mars, each taken at slightly different angles, to form a higher resolution composite. As they explain in a paper published this week in the journal Planetary and Space Science, the program has already combined photos from the Mars Reconnaissance Orbiter's HiRISE camera, which have up to a 25 centimeter resolution, and crafted accurate images with a resolution of a mere five centimeters, meaning each pixel of the new image covers just five centimeters of Martian rock and dirt, an unprecedented level of detail.
"With just eight images, each with a reasonable quality, we can get up to five times more resolution," Muller says.
The duo calls their image restoration technique GPT, short for Gotcha-PDE-TV. The scientists tested and validated their results by creating high resolution maps of the tracks left by the now-defunct Spirit rover and comparing them to the rover's old on-the-ground photos. The researchers believe that GPT images could help to find landing site options for future Mars missions, or even research what's believed to be ultra-salty water deposits under the Martian soil.
"Effectively, with this technique you have drone-eye vision of the surface of Mars," Muller says.
Moving in Stereo
To drastically oversimplify how exactly Muller and Tao's program works: The program can compare the minute differences in pixel placement created by slightly different viewing angles of the same scene. Thanks to those tiny, stereoscopic variations, "we're able to relate the pixels we see in each photo into a higher resolution grid," say Tao. So, what used to be one pixel in an original photo can become a new grid of many pixels in the GPT composite. After coloring in that grid, the GPT program refines the results through several algorithms.
The downside is the GPT image restoration technique requires an immense amount of time and computing power. "For example, if we're talking about a two by one kilometer image, being processed on a 16 core Linux machine with 2.6 GHz and 96 gigs of ram"-that is, a computing workhorse-"well... that one image takes us 3 days to process," Muller says.
A bigger barrier is that there are very few places on the Red Planet that have been repeatedly imaged with powerful cameras from similar angles, the notable exceptions being the paths and landing sites of the various Mars rovers. As you might imagine, the effective of Muller's program falls drastically when you have fewer photos to compare. With six images, at best you can increase your resolution by just 4 times over.
Just a Photoshop Trick?
Muller says the computing time is worth it. He claims his five-centimeter resolution composite images offers a fantastically clearer, sharper image of the Martian environment than can be achieved with anything but on-the-ground rovers. And that could be a boon for studying hard-to reach locations, like the mountainous areas where sub-surface water is believed to be flowing. But Timothy Parker, an imaging specialist and Mars geologist at NASA's Jet Propulsion Laboratory, who was not involved in developing the GPT technique, is a bit skeptical of Muller's five centimeter resolution claim.
"My opinion is that while what we're seeing here might be some increase in resolution, maybe two or two and a half times the original images, the rest of clarity you're getting is from a comparable improvement in sharpening filters and other noise reduction," Parker says. He's basically arguing that while the GPT technique is undoubtedly increasing resolution, some of the sharpness of the images you can see is more like using Photoshop than getting a new zoom lens and actually getting more image data to work with. Jim Bell, a planetary geologist at Arizona State University, who also wasn't involved in developing the GPT technique, isn't in full agreement with Parker's criticism. "Honestly, I think some of this argument could simply boil down to semantics," he says, noting an ongoing scientific disputation on the technical definition of pure image resolution.
"I haven't gone and backtracked through all the math, and obviously people should be concerned about the possibility of these techniques being oversold," he says. But however you define the increased clarity "this technique is clearly better than the original images-there's clearly an improvement," Bell says.
Bell is excited about the increased image clarity, however you define it. "Because geologic processes operate at many different scales, an extra level of detail can help scientists decide between competing theories," he says. For example, on the Martian surface, a channel carved by water or created merely by a dust avalanche might look identical from a certain distance. "But when you zoom in, all that fine scale detail, like the precise curvature of the edges of that channel, or the pebble sizes, that suddenly tells a more complete story," he says.
Clearer photos open up Mars to an entirely new world of scientific questions.
