Video from Artemis II flyby of the Moon will not initially look spectacular
- Thread starter JournalBot
- Start date
Upvote
-6
(1
/
-7)
And there again we see the condescension. Rather than casting aspersions at the mental state required for someone to be willing to confidently contradict you, try introspection, a topic with which I see you must be unfamiliar. Good day!
Upvote
-8
(2
/
-10)
Given your previous reaction, I'm not going to explain you what things like humor, irony, sarcasm are.
Given clearly you never heard about them, I suggest to google yourself about such strange words.
Given clearly you never heard about them, I suggest to google yourself about such strange words.
Upvote
-4
(2
/
-6)
Oh, that's a neat site. On that site, at iso 3200 up to the practical maximum, the D5's dynamic range is above almost anything else they have data for. The Canon R3 may have comparable DR with better signal to noise (less noise at a given iso), and the last couple sony a7's are pretty close in DR, but that site's charts might justify the D5 even today. It does say it applies an adjustment based on pixel size, and I didn't have time to read enough about their methods to be certain whether they're accidentally making lower resolution cameras like the D5 look better than they are, which is common. Some of the pages on that site say different things about what their actual methodology was.
The reason it's possible, depending what they actually did, is that no matter what the sensor-pixel SNR is, once you downscale the original images from different original sizes to a fixed-sized output image, the output SNR gets better as a function of how much you downscaled, because you combine the signal from multiple pixels but the uncorrectable part of noise is random, and when you add the randoms up they often cancel out. If the true SNR changes, so does the point where that SNR intercepts the threshold, and so does the DR that you get using this new intercept value.
Upvote
1
(1
/
0)
He uses a personal definition for photographic (vs engineering) dynamic range, 1:20, which is more practically useful in assessing a modern sensor's ability to resolve luminance for renditions:
https://www.photonstophotos.net/Gen...ngineering_and_Photographic_Dynamic_Range.htm
I find the page very useful for comparing cameras, maybe less so for assessing actual performance in a quantitative fashion.
Upvote
0
(0
/
0)
Quite possible.
I was also kinda dissapointed tho in how much pixelation break-up they had in the stream. Not sure what the point is in having high definition video when it jumbles into massive blocks frequently. You'd have thought they would be making darn sure to have a good link with no packet loss or signal issues in the feeds.
Upvote
1
(1
/
0)
Statistical
Ars Legatus Legionis
The point of HD video (and photos) is that everything captured is coming back to Earth in its full quality. Improving cislunar communications with relay sats and optical links is part of NASA longterm goal but it is going to take years and lots of infrastructure funding.
Realtime 4K highbitrate video streaming from cislunar space without pixelation and other issues is likely beyond NASA capabilities for at least the next half decade.
NASA does have plans for a pretty comprehensive constellation of sats around the moon called LCRNS (Lunar Communication Relay and Navigation Services) which would provide not just high bandwidth communication links but also PNT ("GPS") and SAR capabilities.
How many sats is hard to say because NASA wisely didn't define the number of sats in the contract only the capabilities and coverage however the full network (2032+) will likely require a dozen or more.
Last edited:
Upvote
4
(4
/
0)
There are noise-reduction processes that could have been added to make the effects less jarring. Some of today's ML-based routines do an amazing job of filling in the highest-frequency content when the signal partially degrades. In fact, NASA probably could have trained their own system based on modeled data that would have been very close to the real thing. But that would have required NASA to put anything but the bare minimum of work in their PR efforts. It's the same lack of effort that went into the launch coverage where bandwidth limitations aren't a thing.
This is all ground-side effort. There would have been zero impact on the mission objectives. But this is NASA operating on a shoe-string budget except where the real money is being
Upvote
2
(3
/
-1)
It's fine if it's 1:20 or some other number, the problem is whether or not he accidentally compares cameras unfairly, making higher-resolution cameras score worse than lower-resolution ones. I don't know if that's what he's doing or not.
ML filling in the gaps would be exactly what they shouldn't do, it would be very bad PR to be caught faking some of the media even if they were using legitimate data as their base.
Upvote
5
(6
/
-1)
ML-denoised data is just a representative of the photons that hit the sensor as noise-filled frames. It's all for PR purposes anyways. As such, being more pleasant to consume would be better.
Hell, offer both streams. It's terrestrial at this point. Who cares if you double up the feed? Finally, release the actual video stream in full definition after the optical signal was established again.
Edit: You do realize that all those iPhone photos NASA has released are filled with ML enhancements, right?
Upvote
1
(2
/
-1)
Regarding the DR charts, he does some sort of normalization based on the particular sensor's circle of confusion, a bit beyond my ken. If you're seeing some incongruity, might be there.
Post-processing is a slippery slope, as anything done to make a visible rendition is changing the recorded values, just how digital photography works. You can't have a pleasing rendition from a digital capture without black subtract, white balance, demosaic, gamut transform, and some sort of tone curve, all are damage to the original data. ML noise reduction is no more egregious than the various deconvolution kernel-based algorithms, e.g. Richardson-Lucy, all are intended to reduce statistical noise in the lower, less-resolvable values. RL deconvolution was a NASA initiative, IIRC, designed for putting resolution into faint deep-space captures from the various telescopes. They do color-shifting all the time, as most of their deep space sensors are recording in bands outside the visible range; done to make useful renders for study.
Indictable manipulation would be in the category of making up objects that weren't in the scene.
Upvote
-1
(0
/
-1)
Heh, the HDR, or 'high-dynamic range' of cell phones is a combination of multiple exposures. I find it a bit creepy to take a photo with my phone, immediately look at it in the gallery, and watch as the software replaces the blown-out rendition with a one that properly exposes the blown-out regions.
Upvote
0
(0
/
0)
I disagree with the characterization of those changes as damage. When the purpose is to reproduce what a person would see if there, processing the sensor capture to shift the image towards a correct representation is just part of the purpose of the system. (Of course, it is different if the data being captured in the purpose and the processing is to make it more comprehensible to humans.)
Upvote
2
(2
/
0)
Every single manipulation of a measured luminance from a sensor's photosite departs from the measurement, therefore, 'damage'. The light at the scene is the reference, as it is what enters the eye of the observer.
To make viewable renditions upon a medium, all manner of manipulation is required to meet the medium's limitations. What you look at in a photograph on a monitor, or worse, paper, is just a coarse approximation to what was observed at the scene.
Ref: wrote my own raw processing software when I couldn't find any other program that did it the way I wanted:
https://github.com/butcherg/rawproc
Upvote
0
(1
/
-1)
Raw sensor data isn't valid data. It's device-dependent data. Only device-independent data is valid data.
However, the process of removing instrument-specific artifacts doesn't count as damage. It's necessary processing.
Upvote
1
(1
/
0)
You apparently haven't worked with photographers who insist on editing JPEGS and wonder why they look like shit. Damage was already done...
Edit: speling
Upvote
0
(0
/
0)
Hey, doesn't matter what I know or whether it makes sense. Just matters whether someone can make it a talking point, and it doesn't have to be that strong of one to still work for conspiracy theorists.
Upvote
0
(0
/
0)
At least on the NASA stream we were watching, even the ground based video during ascent was breaking up and pixelating frequently as well. Since it was views looking up from the ground I'd have thought it would be on fiber or similar easy to do high bandwidth terrestrial links. The commentators never broke up so I don't THINK it was on my end
Upvote
0
(0
/
0)
During one dress rehearsal or another, there was some commentary that communications to the launch site is always iffy. I'm guessing all the tall metal structures wreak havoc with micro- and radio waves. Which also probably suggests there isn't a fiber run to the launch site.
Upvote
2
(2
/
0)
Interesting - I would have thought they'd have all their official camera tracking sites run with something very solid (even if its temporary spooled out and rolled up after the event)
Upvote
0
(0
/
0)