Ilumin8 - claiming that upscaling 1080p produces better results than native 4k?

[Sabovic Adis]

Obviously...UHD won't be the only standard...but we may well see an ecosystem where screens can automatically switch btw UHD and 1080 content/standards.

Not too disimilar from how most American BROADCAST CABLE boxes have parallel 1080 and SD options now....

This feels very similar to growth of 1080 HD a decade ago. And today...1080 is is a "standard"...and SD is fading away...

These things clearly work in cycles....

But the number of channels in cable is...weeee, and there's only so much bandwidth. Something's gotta give, and it ain't the number of channels. It's quality. It gets squashed.

Similarly to video, there's compression in audio too - mp3, aac etc., and on a high-res audio system everything gets more audible. Even the flaws.
On a high-res video system everything gets more visible...right? :biggrin:

 

[Nick Morrison]

But the number of channels in cable is...weeee, and there's only so much bandwidth. Something's gotta give, and it ain't the number of channels. It's quality. It gets squashed.

Similarly to video, there's compression in audio too - mp3, aac etc., and on a high-res audio system everything gets more audible. Even the flaws.
On a high-res video system everything gets more visible...right? :biggrin:

As household bandwidth expands, I dont think the bottlenecks you foretell will be so much of an issue. Netflix and Youtube already stream 4K.

However, everyone is entitled to an opinion. If you believe in the gold standard of 1080, and remain skeptical of UHD, thats certainly not an uncommon position.

That said, Arri (the "gold standard" of traditional cinema) was saying the same for almost a decade, and then in the past year did the following:

1) Released UHD upscaling for many of their cameras.

2) Released 6K Arri 65 Imax camera

3) Announced new 6K S35 sensor.

So clearly...even Arri is starting to prepare for a media landscape beyond 1080...

 

[Sabovic Adis]

As household bandwidth expands, I dont think the bottlenecks you foretell will be so much of an issue. Netflix and Youtube already stream 4K.

However, everyone is entitled to an opinion. If you believe in the gold standard of 1080, and remain skeptical of UHD, thats certainly not an uncommon position.

That said, Arri (the "gold standard" of traditional cinema) was saying the same for almost a decade, and then in the past year did the following:

1) Released UHD upscaling for many of their cameras.

2) Released 6K Arri 65 Imax camera

3) Announced new 6K S35 sensor.

So clearly...even Arri is starting to prepare for a media landscape beyond 1080...

Nothing against 4K, 6K or even 8K per se, only against fake ones.
I mean, 1920x1080 8 bit RGB video, gamma corrected @ 2.2 is, well, acceptable. MPEG 2, AVC & VC1 is bare minimum, (but it gets advertised as wow, uh, ah...:lol, then it gets squashed again, and then it should be upresed as suggested by Iluminati?
How about pushing for better infrastructure instead of pushing more junk through it? Programs get better by doing so too.
I mean, if Ilumin8 is really that good, why not use it professionally? In Cinema? Do we really need 250 Mbits/s in movie theaters? And that's for 2K. :biggrin:

Everything' s so saturated...:cryin:

 

[Tom Lowe]

Upscaling is bullshit. If it worked, why would anyone bother to buy a Blu-ray instead of a DVD?

 

[Matt Hoover]

Exactly...just as still photographers using Photoshop have known for a long time, upscaling is not adding new pixel information or detail...it's just interpolation - this is pure marketing BS. Furthermore, this company has world dominating technology but only an email contact? No office space or physical address, much less a phone number? And who are the founders, two guys with no previous history of developing compression algorithms or any relevant work experience or history? It's pretty obvious this is a sham company folks...a fugazi.

 

[Sabovic Adis]

Better than exact!

Better than exact!

Someone must've told them that science is just the best guess...and they took it literally.
:lol:

 

[Wayne Morellini]

Re-edit: Graeme, compression work stuff below compared to redtay. I have a great respect for redray, as it seems a great body of work that can benefit people. My worded comparison merely inidicates that it is far from.the end of the road as you would think from people's reactions. Along the lines of present developments it may seem to be out there are in terms of performance, but in research terms, not compared to what could be achieved. Still, great product, and humbly, much appreciated.

"illumin8’s algorithms leverage adaptive technology that upsamples:icon_bs:, color corrects, de-noises and transforms video in real time. One of the differences between the existing video compression formats and illumine is that illumin8 doesn’t stretch or add pixels; illumin8 uses adanced edge-detection technology to identify and preserve object details within a video to be transferred and displayed at the greatest level of quality that the display device supports."

It makes sense that they are going for the illumin8 version over the native resolution version because they've not just upscaled it but also polished the image in a way that's pleasing to the eye. It just goes to show that there is more to creating a nice image then taking it at a higher resolution.

I have been in bed having a bad turn, so ill try to answer well. Yes, that is right, but what Elsie said to Graeme is right too. I wish Graeme would address it?

They apprently found factal algorithms could predict/simulate substructures. And apart from a few things that I won't get into for private intellectual property in my own work, what you said is so. There are a bunch of things going on I'll try to summarize, as my turn is lifting now.

Hyper realism/super realism. I came up with a theory for an essay in university of abstraction producing heightened recognition in an image which explained the mechanisms of hyoer realism art also. These techniques have been used in digital TV to spice it up to look better than real life. Eliminating niose, and various fine details is one of the hidden techniques, increasing color and contrast as the other two techniques commonly, but deliberatly used. Edge detection accidentally taps into another technique (deniosing, and de detailing being other side products from compression techniques). So yes, upscaling an image and doing this can look heightened compared to a native 4k picture, but doing these things to a 4k picture will make it look better also.

Now stand back, for over a decade I have publically put forwards a technique of recording a lower resolution compressed file along with a difference compression compared to the upscaled version. Thus, if the scaler is good, it will estimate close to the original and so the the recorded difference would be a lot less than if you had just compressed the original 4k file. Now various people now working at red would have seen a shit load of new compression ideas I published back then So not patentable, and I would still like to know how redray actually works.

Going back n further from this, I have come up with further improvements where I hope to record highest grade compressed 8k at 9mb/s+, with very low energy compression. At better than redray compression. That's how I was going to do a 8k ring camera. RR is shit compared to this. If you want something done, ask a true master.

Now the fractal stuff, it was my objective in the mid 1980's to develop a 3D graphics technique that could use fractals firmukar as part of it. 3-5 years ago I finally figured at a good way to do it, and also much better ability then the mid 1980's "dream". A solution that is not just a dream, it is something you put the vitally important time into working out. Sometimes over decades.

Anyway, I have not read into this technique but briefly skipping sampling words in the article during my turn, it seems to be along the lines of my old work.

Now, getting real about this routine. You can only get as much out of it as you can A) estimate (including good guesses) B) As much information as originally there to work with. Bueve me, during university in the mid 1990's U had two engineers trying to get me on the rope, after I observered an information compression law they quoted to me was wrong (I either had just worked out something that contradicted it the previouse week, or on the spot). One is a top international engineer. They couldn't figure out how it could be possible, and retired downstairs sulking. So it is not just some smart asses around here (not you Graeme, you have real ability). Now, this all goes along the lines of my long term work. Such pursuits were popular within compression circles in the late 80's and early 90's

Another hint is video chat compression, achieving thousands to one compression by the late 80's. And if these two engineers had realised this they would have been able to see how the law was wrong, but at that stage not even I regarded this in relation to the laws or my work. They modeled the lios and face in speech, and uses it as a basis to map the actual movements against, greatly increasing the compressibility. Even on your mobile phone the voice codecs coming out in the 80/90's broadly mapped the human voice as a multiple stage wund pipe, thus rendering the voice by model changes etc (why a truck or niosy environment uses to make phone calls drop down into very very low sampled uncompressed samples). This the law to do with no more information from a compressed data stream compared to what was put in was broken again, though I did not know about this at the time. I simply had figured the week before that using a dataset/destination/modelling on the receiving end you could get more information out then the information sent. Just merely dynamically specifying the technique to use on the receiving end great improved compressibility. A useful cheat. Pretty good work to figure out (so quickly) for a niave observer with no real compression or mathematics background (at all). The problem is that the people that make these observations usually have aspergers, which is useful in observations and observing things, not so much in getting around things very innovatively. They tend to get stuck in the observation steadfastly insisting on it as handed down wisdom, rather than figure out ways around it, like my friend was. I have aspergers, so know what that is, but I also have liw latent inhibition from tourettes which means I see many, many observations and abstractions in one time period. I literally will see and rapidly access and pick from long lists of choices and materials and abstractions going through thousands to tens of thousands of things in a sitting, and in accessing the fine grades of amounts in balancing things millions of options in a sitting. Those were the days. So for the ones that have insulted me and the truth everywhere, saying I'm dreaming, I'm the real deal, not their delusionally sour dreaming. That is the tip of it.

 

[Marc Wielage]

That said, Arri (the "gold standard" of traditional cinema) was saying the same for almost a decade, and then in the past year did the following:

1) Released UHD upscaling for many of their cameras.

2) Released 6K Arri 65 Imax camera

3) Announced new 6K S35 sensor.

Small correction: I did a quick web search and have not seen anything to the effect that Arri is definitely working on a 6K pickup. There are a lot of rumors out there but I don't see it confirmed by Arri.

I predicted last year they'd have at least a 4K or 5K version of Alexa at NAB, and I was wrong. Maybe this will be the year. I'd be very surprised if they didn't, but I think it'll be hard to upgrade all the 3.2K Alexas in the field for 6K. Some of them, maybe.

 

[Sabyasachi Patra]

Graeme, some years ago I used a program called Genuine Fractals on digital stills that would blow up a small digital photo to a size suitable for printing and framing as a large print. And while I didn't pixel-peep, to my eye the results were great.

Have you ever checked out how they were able to increase an image in size while seemingly retaining the quality?

I have used Genuine Fractals in the past. It was fine. I haven't checked the latest versions of that software. It is now owned by ON1. I am sure they would have made it better.

 

[M.Halsell]

Will we be having a similar conversation when it comes time to intercut 4K into an 8K project? The acceptable compromises of upscaling, that is?

 

[Elsie N]

Can someone familiar with Pro res comment on upscaling that?... like from 2k to 4k.

 

[Ketch Rossi]

I have used Genuine Fractals in the past. It was fine. I haven't checked the latest versions of that software. It is now owned by ON1. I am sure they would have made it better.

Ah,

Those days GF to make bigger prints out of my DSLR pics... Then I switched to MF...

It works if you can't have or do any better with real resolution, otherwise I firmly belle that nothing can replace the results of true resolution and sensor size.

 

[Wayne Morellini]

Where has all the fractal compression stuff gone, its not on the on1 site?

 

[Ketch Rossi]

Not sure Wayne,

Have stopped using it a long time ago.

 

[Phil Holland]

The hardest truth regarding up scaling and why it works/doesn't work comes down to subtle details. While algorithms are great at producing an informed guess at edge detection, subtle details that are never captured at a lower resolution and thus never seen by the algorithm can never be created. It's a compromise. A comprise that will allow for older content to live on again in modern times. Some productions will have the resources for a proper remastering and that's where the fun is.

 

[Wayne Morellini]