Wikipedia:Reference desk/Archives/Science/2017 December 25

Welcome to the Wikipedia Science Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

December 25

is it possible to upscale old TV series into 4K

Is it possible to use neural network to upscale an old TV series called "six million dollar man" into 4K so that it looks crystal clear on mordern TV set? 110.22.20.252 (talk) 02:51, 25 December 2017 (UTC)

See 4K resolution which says "3840 × 2160". An old TV series is (very approximately) 640*480 480i, (I quote 640 as 4/3 * 480). The aspect ratios are very different (new 16:9, old 4:3 = 12:9). Assuming you want things the right shape and uncropped, you want in the order of 4.5 new linear pixels per old pixel or about 20 new pixels in the area of one old pixel. I have no idea if a neural network would be any help in this task. I guess it is your way of saying "automated" to avoid humans spending lots of time doing the task. A computer could of course be programmed to do the resolution change. Extra information would not authentically be present in those new pixels - it would just be invented by the program's algorithm. Is that ok for you? -- SGBailey (talk) 07:50, 25 December 2017 (UTC)

For still pictures you can't do much better than just filling in the gray values of the newly added pixels using interpolation methods (although more sophisticated methods exist that invoke scale invariance). But in case of video you can use the motion of the objects to implement super-resolution methods. Count Iblis (talk) 15:30, 25 December 2017 (UTC)

Here, conceptually, is a scheme to allow useful resolution enhancements. The basic issue is that a single frame contains no more data than number of pixels*number of light levels*colors. So the single frame can merely be interpolated which adds some amount of fake data, noise, and artefacts. But a series of frames offers much more information, not a new frames worth per frame, but some. So, run the entire dreary episode through, and build a super accurate 3d render of the guy who can't act. Then use the motion of the original episode to guide the motion of the super accurate render, with physics as necessary. Then repeat this for every other object in the episode. It will still be a boring little morality play but it will be at much higher res. If you want to turn 4:3 into 16:9 then given you have a trajectory for each object you could fill in the edges of the screen. So then it will be a boring little morality play at high res in widescreen. Then you need to do the same with the audio track. I suggest the easiest way would be to synthesize the voice of each object rather than trying to muck about with the original audio. As to neural networks, they are a mere detail, not the solution.Greglocock (talk) 22:22, 25 December 2017 (UTC)

• I'm not the OP, but I have a follow-up question. Assuming that the series was originally recorded onto film or video (I don't see it specified in our article), wouldn't the analog nature of those media mean that the episodes do not have a set resolution (as SGBailey implies)? Now, some analog film is crap regardless, but it's not really a resolution problem per see, is it? The film could be too dark or too murky or out of focus or whatever, of course. Matt Deres (talk) 04:21, 27 December 2017 (UTC)

Photographic film has a "resolution" too! It's called Film grain. --Kharon (talk) 05:34, 27 December 2017 (UTC)

Note that the vertical resolution of an analog television video signal is fixed at recording—it's a raster scanned image, with a specific number of rows. In principle, one might be able to extract additional horizontal resolution from the continuous analog signal across each row. In practice, there's a physical limit there too, since there's an upper limit to the signal frequency you can record on tape using helical scan recording. A good commercial recording system (and tape) might be somewhat over-engineered, but it strikes me as unlikely to have the desired four-fold resolution improvement hidden away. TenOfAllTrades(talk) 21:22, 27 December 2017 (UTC)

Time travel

The classical concept of time travel is where one goes back and changes the past (and thus also the present). Hence the Grandfather paradox and the question why we aren't having guests from the future right now. But what about some other type of time travel where one can only SEE the past -- maybe like being there or watching it like a movie or whatever -- without being able to INTERVENE in it? I find this more likely to some day be achieved by science, and yet I've almost never heard of such a concept. Has no scientist ever thought of it? --Qnowledge (talk) 08:07, 25 December 2017 (UTC)

See http://www.sf-encyclopedia.com/entry/time_viewer -- SGBailey (talk) 09:21, 25 December 2017 (UTC)

This is easily achieved by going to the top of a skyscraper and looking down. If the elevator moved at superluminal velocity you will be able to see yourself entering the building. As the speed of light is not infinite, everything you see, including the tip of your nose, is an observation of the past. See also relativity of simultaneity et al. --Cookatoo.ergo.ZooM (talk) 09:50, 25 December 2017 (UTC)

Well... you see the present as far as the photons are concerned, because moving at the speed of light their light cone is tilted... ;) And you don't see your past because you weren't there (whereas, the photons were).

Recommend E for Effort, though there are a number of similar tales. Note that at least conceptually it is possible with present physics if, say, you watched the image of the Earth wrapped around a black hole by gravitational lensing, but also totally absurd (not one photon would make it, I would guess. Though at that point they might do images with might-be-photons, or half a dozen other handwavey explanations. Wnt (talk) 12:48, 25 December 2017 (UTC)

There is no problem with this. Just put a mirror at a large distance and you will see the past in it. Ruslik_Zero 18:21, 25 December 2017 (UTC)

While some of the comments above have already touched upon this fact, I think the point should be more explicit for the OP: the observations in question are, by definition, not time travel, but indeed completely consistent with the standard definition of what an observation is. We have many methodologies, predicated in many different physical mechanisms, for viewing the past state of some portion of the universe, and they vary quite substantially in scope. The oldest galaxies detected by the Hubble Ultra-Deep Field telescope, for example, are about 13 billion light years away, meaning we observe they existed a "mere" few hundred million years after the big bang; in terms of time span, it doesn't get much more impressive than that. But it's also an artificial and limiting distinction to constrain our definition of observation to a discrete measurement of light or other radiation. In reality, any time you deduce a past state of any portion of the observable universe, from the present state of the universe, by means of any reliable process, it is just as much a an effort at "viewing the past" as any other method. Presumably what the OP is really getting at here is that they want a "view-screen" that would allow them to "look" at an exact moment of time (in the distant past) in an exact place. That is probably not possible, for numerous reasons; with a vast enough amount of processing power and enough starting data (from the present point in time) one can create a reasonably accurate replication of a recent event, but the farther back in time one goes, the more quickly the demands of the reconstruction rise, exponentially. But what even present day observational techniques can accomplish, over astronomical distances, is quite impressive in its own right. Snow ^{let's rap} 01:39, 26 December 2017 (UTC)

Thanks, interesting. But isn't it so that nothing can move faster than light? And how are the skyscraper and mirror things (described by Cookatoo.ergo.ZooM and Ruslik0) supposed to work then? --Qnowledge (talk) 07:07, 26 December 2017 (UTC)

Your frame of reference is always at the tip of the light cone. If you send a signal that is reflected back to you, it's limited by the speed of light. The round trip distance is the "past." It's like viewing the surface of the sun and describing it as 8 minutes in the past because it is 8 light-minutes distant. The concept is more complicated if you compare to frames that move relative to each as simultaneity is not absolute, but relative. See light cone to see how the past and distance are intertwined. --DHeyward (talk) 07:29, 26 December 2017 (UTC)

Alleged moon phenomenon of 1882

From Scientific American, 1882:

A Curious Appearance Of The Moon.—A singular appearance of the moon was observed by several residents of Lebanon, Conn., on the evening of July 3. The moon, almost full, was about three-quarters of an hour high. An observer says,— "Two pyramidal luminous protuberances appeared on the moon's upper limb. They were not large, but gave the moon a look strikingly like that of a horned owl or the head of an English bull terrier. These points were a little darker than the rest of the moon's face. They slowly faded away a few moments after their appearance, the one on the right and south-easterly quarter disappearing first. About three minutes after their dissappearance two black triangular notches were seen on the edge of the lower half of the moon. These points gradually moved towards each other along the moon's edge, and seemed to be cutting off or obliterating nearly a quarter of its surface, until they finally met, when the moon's face instantly assumed its normal appearance. When the notches were nearing each other the part of the moon seen between them was in the form of a dove's tail."

Assuming this account is true and not wildly exaggerated, what could have caused the phenomenon described? 2602:304:B041:9A9:94A2:B15B:CF4E:2BF6 (talk) 22:21, 25 December 2017 (UTC)

There are three main possibilities. The first would be a change to the Moon itself, which would be spectacular, but would be visible to many others beyond Lebanon, CT. The second, which appeals to wishful thinking, is that it was something distant from the surface of the Earth so that multiple points in Lebanon, CT would see it align with the Moon. The third, which being most unglamourous seems most likely, is that the multiple observers in Lebanon, CT were standing together, allowing them to corroborate one another's accounts but not giving any real perspective at all. That would suggest some kind of "optical phenomenon", of which there are many, all strange and spectacular, but not very uncommon. Wnt (talk) 00:58, 26 December 2017 (UTC)

Possibly relevant article: Transient lunar phenomenon. Also in 1822, (according to this questionable source): In 1822 German astronomer Franz von Paula Gruithuisen announced he had discovered a "lunar city" possessing "dark gigantic ramparts". —2606:A000:4C0C:E200:F89E:4736:5756:93D4 (talk) 02:30, 26 December 2017 (UTC)

The basic story is quite similar in its generalities to the story related at Giordano_Bruno_(crater)#Formation. Matt Deres (talk) 14:09, 27 December 2017 (UTC)