DSP brings you a high-definition moon walk
The National Aeronautics and Space Administration recently discovered that someone at the agency recorded over the master copy of the video of Neil Armstrong's famous Apollo 11 walk on the moon. One small step for one man succumbed to one major error by another. Surviving, however, are myriad transcoded copies, edited versions, and other variants on the original recording. None has the information content of the original 1-in. master tape, which NASA says has been degaussed, recertified, and reused—no doubt for something more historic.
Nonetheless, John Lowry, founder of Lowry Digital, reasoned that the video might hold enough information to re-create the original recording and maybe even to interpolate it to high definition. Lowry's company has developed the necessary image-enhancement technology, having initially developed it for use on later Apollo missions. This new project slightly differs, however. This time, Lowry is working primarily from four sources with assistance from a variety of other recordings. Apollo 11 transmitted the original video back to Earth on a 10-frame/sec, 320-line, slow-scan downlink. A number of recordings exist in which scientists have rescanned and frame-shifted this source into PAL (phase-alternating-line) or NTSC (National Television System Committee) formats. One other source that is the only option for some of the material is a film from a spring-driven, 8-mm, 16-frame/sec movie camera that someone at Mission Control aimed at a TV monitor during the mission.
The restoration will primarily use Lowry Digital's temporal image processing, which processes long sequences of frames to remove transient noise. Scientists at the company have also added corrections for problems specific to this material, including vignetting, limited dynamic range and ghosting in the imaging tube in the original RCA camera, time distortions during format conversions, and noise from media aging.
Lowry is reportedly using Nvidia Tesla GPUs (graphics-processing units) programmed in the company's CUDA (Compute Unified Device Architecture) to implement the algorithms. Nvidia claims that the GPUs are approximately two orders of magnitude faster than CPU computations, reducing the processing time to less than one minute per frame.