High Definition-A brief Discussion

[GodofaGap]

Many ABX tests are flawed and many people simply lost the confidence. It's so easy to trick an ABX test, I said it too many times.
Oh so that must make it true then. :)

The only time an ABX test needs to be performed is when someone claims to hear the difference and then it needs to be performed by that person. So unless he wants to trick himself, there are no flawed results from this. Absence of proof is not proof, but you'd hope the critics of digital audio in general and lossy audio in particular would bring more to the table then they are doing now (which is mostly nothing). There is a test that can show they hear the differences they say they hear, but yet those people are often the ones not willing to use it.

Of course, the inevitable conclusion is that mathematics doesn't work and therefor this whole conversation never existed.

Poof. There goes the internet. Poof. There goes my computer, there goes my TV, there goes my radio and the rest of the electronics devices.

Yes, I'm sure I'm the idiot here coz it is starting to look like 1864.

Poof.

[Ghitulescu]

The only time an ABX test needs to be performed is when someone claims to hear the difference and then it needs to be performed by that person. So unless he wants to trick himself, there are no flawed results from this.That's why I said this.
The only real ABX tests are done internally in a company and they never get public.
Which is equally true for any person. S/He will not trick her-/himself because s/he has a legitimate interest. One can even compare incomparable things if s/he has an interest in comparing them.

Like it was once said: The music is in the ear of the listener.

[jmac698]

Monao,
Actually you're right! But they did mention the idea of perception vs technical standards, which is how I took it as an alternative thread to his question

for me second Scenario is considered as better Quality than first

And yes I agree, HD is defined in the standard as a resolution, has nothing to do with quality. How could I disagree? It's a published standard.
I'm still making a point however about the marketing use of HD as being perceived by the layperson as equating higher quality. Then I'm looking at, is there a way to translate higher quality back into a technical standard?
That is, is it technically possible to define quality somehow?

I think as technology progresses, there will be nothing but 'retinal' displays and ultra high frame rates and full color gamut. No doubt we will reach perfection in the mainstream in even 10 years (the UltraHD standard is progressing, we now have 10 layer Bluray Discs on sale and drives to use them).

[ramicio]

Your forensic image processing still uses two pictures and other known data, so it has nothing to do with this case of video upscaling. If the camera was black and white they can still possibly get the color of the car if there are eye witnesses. We already have retinal displays...it's called proper viewing distance.

[jmac698]

Just to clarify, my post where I reconstructed the letter "A", did not need any additional information. Yes, it does apply to upscaling.
Please visit http://www.compression.ru/video/super_resolution/super_resolution_en.html
I think this technology is used in some Samsung products.

[Ghitulescu]

And yes I agree, HD is defined in the standard as a resolution, has nothing to do with quality.

It allows a higher optical resolution to be stored within. It doesn't require, just provides the technical means.

[ramicio]

Your example used 2 frames to construct one. If that's not additional information then I don't know what is...

[jmac698]

We're talking about video here - why is it not fair to extract extra resolution from different frames? If done in realtime, there will be a slight lag in playback,but it will be upscaled.
Many upscaling algorithms use one frame, but there's no reason why they can't use more.

I also gave you two examples of how it's possible to create accurate detail from just one frame; the fractal example and the "smart" brain example (which uses neural nets, training previously on HD videos).

[ramicio]

It's not accurate detail, it's just enough detail to be able to extract a plate number. It's still interpolated data. So I guess in your world the frame interpolation tech. in TVs is also just as good as filming in those rates in the first place.

[jmac698]

No, not at all. In a previous message I listed 3 scenarios of upscaling; superresolution upscaling comes closest to the real thing and has a significant measured accuracy (not perfect of course). If you looked at the link, you will see that the quality is quite a bit improved. So we do agree after all!
I'm glad you've learned more about upscaling techniques.

[ramicio]

No, I don't agree. I'm not one who add an avisynth script into my player for upscaling. I make sure the source is up to snuff, not compensate for space/quality and then use post-processing to give me a fake.

[Ghitulescu]

Without said upscaler the playback would be even more grotesque.

[ramicio]

Which means it's gross in the first place because of upscaling...

It's the same garbage they spew when they upscale SD content to HD and argue that it's HD. A good example would be the Family Guy version of Blue Harvest on Blu-ray. It wasn't made in HD so they upscaled it to HD, and with line art it's disgusting. Why they can't just go re-render it is beyond me.

[jmac698]

I don't either, I encode from HD to anamorphic xvid, that I have to use for compatibility reasons. My bluray player does upscaling anyhow, I don't have a choice! And I believe ppl use upscaling scripts under ffdshow to view their SD clips for better looks (on the computer).

If the only source of a movie is SD (not all are released in bluray), then why not?

[ramicio]

There is no reason to have to compromise in compatibility other than cheapness. I don't know of any movie before the HD era that wasn't shot with film, so they ARE in HD. If they didn't use a film camera then it probably was some low-budget piece of crap that I wouldn't be watching anyway.

[jmac698]

So did we pay for the introductory argument or the full course? :)

[jmac698]

Yes, most earlier material is in film, but some TV shows moved to video tape later.
My point was that this film based material is not released in any HD version yet. In some cases the film is lost, like with Dr Who (and yes it's low budget, but fans still love it :)
My encodes must be viewed on devices which only support Divx/Xvid (my dvd player, mediabox, and portable device, and low-power netbook which can't play HD), which I started before bluray existed.

[ramicio]

Exactly...cheaping out on playback devices. Even on tape it wasn't some 480-line crap, it was closer to common HD than you think. How is it so hard to understand that upscaled crap is not HD?

[jmac698]

On tape it was 486 lines in NTSC, and even betacam SP is not the same quality as a modern DVD.
And I've said a few times now, I agree with you, upscaling is not the same as the original produced in HD.
I'm not cheaping out on display devices - I have to start a new collection in HD now that Blurays are available, and that will take some time.

[ramicio]

Rent blu-ray from netflix and rip them...

[jmac698]

netlfix ? What's that? Is that some US thing?

[ramicio]

It's in the US and Canada. It's mainly a streaming service, and they do mail rental too.

[jmac698]

Interesting use of superresolution, just as I described with the letter "A":

Hasselblad has announced a camera capable of producing 200MP images based on shifting its 50MP sensor. The H4D-200MS uses an extension of the company's sensor-shift, multi-shot (MS) technology to create a 200 megapixel file from six images taken at slight offsets. It can also use the four-shot mode used by the H4D-50MS that shifts the sensor by one pixel in each direction to capture all colors at each position. The latest approach adds 1/2pixel offsets to increase the captured resolution to 200MP. The H4D-200MS is available for €32,000 or as a €7,000 factory-return upgrade for H4D-50MS owners

[ramicio]

Still not real. For it to be real it has to be one image. They are capturing different moments in time. Good luck with motion with that.

[jmac698]

Ah, but there is no real - there is a problem called rolling shutter. The camera "eye" is read a row at a time, and there was motion meanwhile.
http://en.wikipedia.org/wiki/Rolling_shutter
This problem exists in most cameras and camcorders today.

Another problem is motion blur, where the "eye" is accumulating light while an object is moving. It's nearly impossible to capture an exact moment in time, because there is so little time to accumulate light. For example, in the clips with high speed photography (slow motion), these need a huge amount of light.

For this Hasselbad camera, there is no information in the press release about how quickly the process occurs, but you will have the same problems as all cameras, if you are taking a picture of sports or other action.

For studio shots, landscapes, this could be a useful feature.

Actually, it is impossible to capture a moment in time, due the Heisenburg Uncertainty Principle. This says you can't measure position at the same time as something else (which can be converted into various statements). I would guess if you applied the formula to a camera, you couldn't record the brightness of the pixels at one moment of tiime?

Are you planning to photograph the bullet train? In practical movements with humans, it's not a big problem.

[ramicio]

Which only applies to CMOS, which is crap. It does not apply to CCD or film. That camera is CCD. Why anyone would use a CMOS image sensor in anything better a camera phone is beyond me.

[smok3]

actually CCD sensors are usually used in phone cameras.

[jmac698]

There are a few advantages of CMOS; they are smaller, can integrate image processing circuitry in the same chip, and faster at aquiring images. But yes, generally a high end camera would use 3 CCD's; for each color.

Since this Hasselbad uses CCD, you would agree then that this superresolution is not as much as a problem.

I think you only need a B/W camera though, as that's how you tend to see things :)

[ramicio]

What the hell cell phone uses CCD? I'd say 99% of phones use CMOS. And that extra 1% would be phones that are otherwise crap. Small sensor size is only a good thing for small devices, such as phones. CMOS are not used in anything serious. If so, then shame on any professional for buying that product. One sensor for each color is still reading the data at the same point in time... But those cameras are extremely high end and pointless, for video, that is. What the hell is the point of anything beyond 4k? For that Hasselbad crap, I don't agree that it's not a problem. You still must take multiple images, which means two different points in time.

[jmac698]

Well, gigapixel images are used in art exhibits:
http://upcoming.yahoo.com/event/8047921/PA/Pittsburgh/Gigapixel-Imaging-for-Science/Carnegie-Museum-of-Natural-History

You can enjoy them online:
http://gigapixelphotography.com/image-gallery/
Tell me those are not beautiful pictures!

And the quality of IMAX is beyond 4k.

Baraka was filmed in 70mm and scanned in 8k:
http://en.wikipedia.org/wiki/Baraka_%28film%29#2008_remastered_release_and_sequel
I've rented it, it looks beautiful!

Having 3 CCD's is not pointless for video, have you heard of the Bayer Filter?
http://en.wikipedia.org/wiki/Bayer_filter

Even witha high end CCD camera, the quality has issues due to the demosaicing process - that's why you hear this talk about RAW processing. One of the processes in raw processing is to find the optimal demosaicing settings.
All cameras, CMOS and CCD, (except Foveon, 3 CCD and greyscale) use upscaling as part of the process, in the form of demosaicing.

From an idealistic point of view, perhaps you might like the Foveon sensor?
http://en.wikipedia.org/wiki/Foveon_X3_sensor#Comparison_to_Bayer_filter_sensors_.E2.80.93_operational_differences

Of course they built in on CMOS, so you will probably hate it. Just face it, there is no perfect anything - there is always a tradeoff. Trust me, engineers dream of perfection as well and try as hard as they can, and thus technology keeps advancing.

Having higher than needed resolution gives you some flexibility in editing - for example you could crop out a mistake such as a boom mike in the frame, while still retaining sufficient quality for your delivery format.
The producers on Baraka said in the making of documentary that the 8k process made a noticeable difference to the Blurray delivery version.

[ramicio]

I see film as perfection. The only thing is the framerate is pathetic. There's no excuse to have any video below 60 Hz. 8k won't show you any quality over 2k or 4k if you are watching in a consumer format. CMOS is crap, period. You can rant all you want about filters and masks, but it's only interpolated detail, not real. If you have 3 sensors, R, G, and B, and each are 1920x1080, then you will get a 1920x1080 image. Anything above that is only some approximation, NOT real useful data.

[jmac698]

Btw, BBC requires 3CCD cameras to be used in programs submitted for transmission on TV:
http://www.bbc.co.uk/commissioning/tv/production/delivery/hd-production-delivery.shtml

Yes, I agree, and all high-end CCD (still) cameras have interpolation, which is as you say crap. There are film cameras that record up to a million fps, but remember film and it's processing is very expensive, which is why it uses the lowest framerate possible. The manufacturers and processors are also shutting down. I think technicolor was one of the best processes, but I'm not a film expert.
Anyhow some films have faded, even recent ones, and had to be restored, like Apocolypse Now.

Film is not perfection, as it requires a lot of light, and quite grainy. CCD can achieve up to 90% light efficiency, film is around 2%. See, there is always a trade-off! The new RED digital camera far exceeds the quality of film. Several new movies are in production using the latest generatation of UltraHD cameras, but you won't get to see them for another year at least. They will be virtually noiseless and crystal clear.

As for, why does Baraka look better on bluray from the 8k scan? Because a down-sampled scan from 8k to 2k is better quality than a 2k scan. This implies that a 2k scan is not an ideal 2k scan. In engineering, in general, "oversampling" creates a better signal. The reason is that it's hard to engineer ideal processing "close to the edge", so downsampling a higher resolution will always give a slightly better picture/audio.

[ramicio]

Good for the BBC, it's not my fault those people are snobs. It makes no point. Interpolation IS crap. It's fake data. This is very dangerous ground saying oversampling creates something better. Better not say anything subjective like that to anyone from hydrogenaudio.org. You don't think it possibly looks better because it came from 70mm film and that's why it needed a 8k scan? Even so I doubt a 4k scan and a 8k scan will look any different in 1920 width.

[jmac698]

Why do you pick 60Hz as the cutoff? There seems to be an internet myth that your eye can only see 60Hz. That's not true. If you have an old film camera, set the shutter to 1/1000s and watch it flick under bright light. I can see it move.
In experiments, for example we can distinguish between a circle and a donut in 1/1000s. Our eyes process different kinds of information at a different speeds.
I think the 60hz myth comes from the (I'll have to look this up later..) but the eye retains an image for some amount of time, I mean that as images come in, they merge with previous images, until the oldest images fade away. We can still see faster motion, just that it's super-imposed on top of older images. I hope that made some sense.

Anyhow, maybe we want to refer to the smooth perception of motion - or maybe we want to be able to do smooth slow motion at home as a special effect on playback? It depends for what purpose you are thinking.

[ramicio]

I used it as a minimum. Your eyes cannot see 1000 Hz. Just because you can see a single flash of light that lasts 1/1000 of a second does not mean you can see 1000 images per second. You would be able to see a bullet fly. 60 Hz just makes sense. 72 FPS was studied to be the greatest point to which beyond that there are no benefits. 72 FPS makes no sense except it's a multiple of 24. Why film makers feel the need to stick with 24's is beyond me. It's just retarded. 48 FPS or 72 FPS would never play on any known TV smoothly (maybe CRT, but they don't count for anything anymore, they're dead.)

[jmac698]

This is my last message, I actually have to work :)
In sampling anything, video or audio, you have to obey the sampling theorem. This says that you get distortions (called aliasing) if you try to record frequecies too high for your sampling rate. If your sampling rate is 44100hz (or 1920 pixels), you can only record frequencies up to 22050Hz (or 960 line pairs? I don't know the unit of measurement here). To avoid this distortion, we first filter the signal. In audio that can be done with a circuit with capacitors and resistors. There's an equivalent in optics, the antialiasing filter.
We can't engineer these filters perfectly, to let in all frequencies just below the cutoff, while perfectly blocking all the frequencies above (that would be called the theoretical "brick-wall filter"). That's because of the physical precision of making analog components.
However if you boost the sample rate, now your filter can be much less precise. So we over-engineer the sample rate and make an average analog filter.
For example, 192khz sample rate and a gradual filter which cuts off at 96khz. Then we downsample to 48Khz (say) and have a nearly ideal 0-22Khz signal.
Hope that made some sense.

http://en.wikipedia.org/wiki/Oversampling#Motivation
You won't understand some of the technical parts of the article, but you will see quite clearly that it follows what I'm saying.

[ramicio]

Talking about that Nyquist crap has nothing to do with video. It's about analog and digital conversions. And its hugely presumptuous. The example of audio and the sample rate being 2x the highest possible frequency has nothing to do with scanning video. It's just a theorem, which means it's only proven on paper and doesn't apply in reality. I understand plenty, I just realize that not everything in the world has to be as technical as it is. Over-engineering to get a better result from false data never works. It only pleases the masses, not people who actually care about quality.

[jmac698]

http://en.wikipedia.org/wiki/Persistence_of_vision

There it is.

So tell me, how can you catch a baseball at 90mph if you can only see in 60hz?
It's going to take me too long to find the reference to the experiment showing we can recognize some images at 1/1000s. But yes, seeing a flash of lighting could be explained by the persistence of vision of the flash on the eye.

[jmac698]

If you don't believe Nyquist applies to video, we have no common reference to continue that line of thought. Nyquist applies to any sampled signal. It's a technical topic you may not understand - I don't claim to be an expert, but I'm pretty sure I understand it, I'm a programmer in DSP and do electronics so I've used these "theorems" in a practical sense, and done a lot of experiments to understand it better.

nice chat ;) I hope you've become a little more informed, at least.

[ramicio]

Flash 2 different images right after one another at 1ms each and tell me that you see 2 different frames. No, they appear as 1 image. The retinal cells can't react fast enough and the bandwidth to the brain is not sufficient. Your example of the 90 mph baseball makes no sense whatsoever. It makes you look like you just like to throw numbers around, list some criteria, and try to embarrass people. A 90 mph pitch takes about a half a second to reach home plate. That's about 30 frames. Again, I did not say the eye sees only in 60 Hz. It doesn't really have a refresh rate. I merely mentioned it as a good refresh rate for video in today's technology. You're just grasping with some arbitrary numbering to make it seem like it's not possible. So if a baseball moves at say 90 millimeters per second you cannot display that at 60 Hz? The viewing angle makes a difference. If you focus on a 1 inch spot then no, you will most likely only see a baseball in 1 frame of a 60 Hz video. If you focus to 100 feet you will see that motion.

[2Bdecided]

Many ABX tests are flawed and many people simply lost the confidence. It's so easy to trick an ABX test, I said it too many times. The only real ABX tests are done internally in a company and they never get public.
Hardly. BS.1116 (double-blind ABX-like tests) are routinely carried out for each new generation of codecs, and the results published widely. e.g. http://mp3-tech.org/programmer/docs/w2006.zip

Cheers,
David.

[2Bdecided]

I used it as a minimum. Your eyes cannot see 1000 Hz. Just because you can see a single flash of light that lasts 1/1000 of a second does not mean you can see 1000 images per second. You would be able to see a bullet fly. 60 Hz just makes sense. 72 FPS was studied to be the greatest point to which beyond that there are no benefits.Not true...
http://downloads.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP169.pdf
...but I suppose those people are idiots too? I suppose you're the only one who is not an idiot?

Cheers,
David.
(Hydrogenaudio regular ;) )

[ramicio]

Is an ABX test now magically also a lie detector?

[smok3]

The high frame-rate can be regarded as temporal oversampling in this context, which
leads to the exciting possibility of allowing temporal aspects of the video such as shuttering and
motion blur to be adjusted by the director in post-production, perhaps even selectively within the
frame. This would add additional “looks” to the director’s palette, complementing the well-known
“video” and “film” looks.

interesting, i want a post shutter-speed_fader please.

[GodofaGap]

Is an ABX test now magically also a lie detector?
Lies are not a problem, because I can't lie for you. If you claim to hear a difference, what then would be rationale for you to lie in a subsequent ABX test?

Still I think it's great we can all have a conversation via medium that according to your own words can't exist. :)

[2Bdecided]

If you claim to hear a difference, what then would be rationale for you to lie in a subsequent ABX test?Not much - but you can lie entirely - saying you've taken+passed a test, and posting the results log to prove it - when really you just typed the "log" yourself.

There are things to catch this in the official HA listening tests (checksums in the logs etc), but it's always possible to cheat if you really want to.


Anyway, this isn't the point of ABX. The point is to remove placebo. Placebo isn't lying.

However, placebo is the basis of much of the high end audio industry, which is why some people get so touchy about ABX.

Cheers,
David.

[ramicio]

And it's easy to cheat an ABX test even if you DO hear a difference. It's called just giving random results...

[madhatter300871]

Ramicio.

I have a wonderful garden, full off grass and flowers. I love sitting out there when the sun is shining and a slight breeze is in the air.... maybe be a little cheecky and pour myself a gin and tonic ... don't tell any one but sometimes before 4pm as well !

Anyway, back to this video stuff ... my hobby .... chill out man, life is too short ! You do seem to be rather aggressive in your view of this topic. Shall I send you my gin and tonic secret recipe ?

I'm still amused at the snobs at the BBC ... what where they thinking ! And as for Nyquist Theorem being crap .... well, beat me with a kipper until Woolies re-opens but that has had me chuckling all day.

I love forums.

[ramicio]

Life may seem short for everyone else...for people who choose to pursue non-material and emotional family BS...but I choose materialism and no people, so life is plenty long for me. Where would the world and science be NOW if people in the past hadn't been so passionate about what they believed. Only in today's fluffy foo foo liberal world are people taught to deem celebrities' realities as their own and not have convictions or their own views.

[madhatter300871]

So why be a part of this forum if you choose no people ? Do you need a valium ?