How artificial intelligence gave new life to old Apollo footage

Apollo's New Moon executive producer Thomas Lucas discusses the process of upscaling NASA images to 4K

Jenny Priestley 15 mins ago

Apollo's New Moon executive producer Thomas Lucas discusses the process of upscaling NASA images to 4K

target=_blank title=Share on LinkedIn class=share-linkedin> Apollo's New Moon executive producer Thomas Lucas discusses the process of upscaling NASA images to 4K

https://www.tvbeurope.com/features/how-artificial-intelligence-gave-new-life-to-old-apollo-footage title=Share via Email class=share-email>

As the world celebrates the 50th anniversary of man's first steps on the surface of the Moon, streaming service MagellanTV has employed artificial intelligence to restore images and footage from the Apollo missions. The digitally restored content has been used in the documentary Apollo's New Moon, which focuses on Apollo's scientific elements of the space programme and how it revolutionised scientists' understanding of the history of the solar system.

In late 2018, the team at MagellanTV discovered software company Topaz Labs who had developed software called Gigapixel AI, which can be used to up-res still photographs. The software employs machine learning techniques to perfect the process.

Once the production team saw what Gigapixel was capable of, they got to work on processing NASA's old Apollo missions footage, as MagellanTV co-founder and Apollo's New Moon executive producer Thomas Lucas explains: With a lot of this old Apollo footage we had to just knuckle down and input it through the processes that it needed in order to be presentable by our standards, which is high-resolution 4K. If you look at the film, you'll see this old film footage has very little film grain in it, the images are quite sharp and beautiful and it looks like they were shot in 4K.

A traditional up-resing tool might look at the detail within a scene and try to smooth out the pixels in between points of detail, continues Lucas. What Gigapixel AI does is try to interpret detail. So when it sees a star in the night sky, if that star is blending in to multiple pixels around it, the programme will tighten it up into a single dot to get rid of the blur, but also it will extrapolate detail, which means all the edges get sharpened.

It works great with Moon images, because there's a lot of detail in the ground as it's all kind of sand-like and dust, so the programme sharpens it all. It eliminates the mushiness that you can get from just blowing up an old image.

The 50-minute documentary features archive film footage that has been blown up from standard resolution to 4K. In fact, it's bigger than 4K, because it was originally all 4:3. So it's the old film formats that were phased out in the 1990s, explains Lucas. We had to get them into a 16:9 4K frame, they had to be even larger than 4K.

Around 50 per cent of the film is made up of actual footage from the Apollo missions, as well as still photographs which were shot on the lunar surface, which were also sub-4K, meaning they also had to be processed. Adding the two together means about 75 per cent of the film is Apollo era NASA material.

How involved was the space agency in the creation of the film? NASA wasn't involved with this film, says Lucas. We're just putting it out there and we'll let people come to us if they have any reactions or any comments. It looks really good, so I think that's what's going to really impress the viewers.

Over the last decade or so NASA has gone back and re-transferred some of the film footage, in particular the camera magazines that came off the spacecraft. They transferred a lot of it to 720p, which Lucas says is what's enabled MagellanTV to produce their documentary.

Some 10 years ago 720p was considered HD, and it kind of still is, but filmmakers like me have moved way beyond that. We want 4K, and even larger. We used other material that was 480p and in fact some of the Saturn Five rocket footage was even smaller, he continues. We scaled up our computing operation and bought computers that had much bigger crunching power. We also got the highest-end GPUs we could find. Even then I think it was taking something like 30 seconds per frame, all told, with all of our processes.

Lucas reports the processing of all the content took months. This sort of thing you can just set it and let it run for two days, so it's not that bad, he laughs. Of course, you don't just process what's in the final film, you have to do much more than that, to get all the correct material that fits together, and you can edit with. It took a number of months to have this and of course, the film had to evolve as well through the editorial process.

LINK:	https://www.tvbeurope.com/features/how-artificial-intelligence-gave-ne...
	See more stories from tvb

Most recent headlines