4k by 2k or Quad HD...lots of rumors? thoughts? - Page 30

Onkyo has announced four new AV receivers with the TX-NR616 and TX-NR515 supporting video upscaling/output at 4K. After looking for more information I found the TX-NR616 datasheet which notes that pass-through of 4K video isn't supported. This sounds like a bad idea to me since it could cause consumer backlash if people buy a "4K upscaler" AV receiver only to discover later on that it can't actually pass-through 4K video. Onkyo didn't hide this information but not many people actually read the datasheet of a product and I currently see no mention of this issue on the Onkyo product page.

Maybe they can fix with a firmware update? For the few that have both the Sony 4K projector and the Onkyo 4K receiver, it is no contest. No surprise the upconverter in the $20K Sony looks much better than the upconverter in the Onkyo.

They really need a pass through for 4K because in native resolution the displays are stunning. Don't know if I can go back to 1080p

OK guys, the issue is not if the diff btw 4K and 2K is visible. The issue is if it is visible from the typical HD viewing distance with same compression and processing applied to both formats.

While those native 4K shots look very good, as native 4K is suppose to look, i see nothing that makes me wish for it to come sooner.


I have a feeling compressed 4K to a BD is going to look different from all these demos we're seeing. Where it's more limited to space and bitrate. Then the cry for 8K will begin.

Star Wars has black bars, looks like SD, are you shure its the 1080p bluray? The upper picture is full screen, might be a photo, looks very nice.

Its nice seeing those pics but there is not enough info.

I don't know how anyone thinks they can come to any useful conclusion about 4k vs 2k when the conclusion is based on a comparison of a still of a girl in front of bamboo at 4k vs a single shot from an action movie with a fair amount of cgi, then taken with a camera off of a display where the image itself is not filling the full frame, then downrezzed to 1024 pixels wide, then judged on your computer monitor.

^I'm sure ppl will be soon showing close-ups of pixel matrix to prove there is more res and details in 4K than in 2K.

^ I'm sure you'll be still be telling us it doesn't matter, we can never see the detail unless we sit with the TV mounted to our foreheads.

But in the case of 1080p -> 4K we wouldn't have to re-sample because the 4K resolution we're talking about is exactly double 1920 x 1080 in both directions... so the original sample points can stay in tact with just new information sampled in between. it's even-integer upsampling and should look great and with no downside if done properly.

Amen.

Unless someone is showing a closeup of a small screen area so that the PC monitor resolution far exceeds the native resolution of the portion of the image being captured. But so far that's not the case... and I find it humorous that folks are looking at small pictures on their less-than-1080p computer screens trying to draw conclusions of image-quality comparisons between pictures of 4K and 2K televisions...

People shouldn't assume that just because this is an internet forum, that we're all viewing on a computer monitor. Or that our computer monitors don't have a resolution beyond 1080p.

I agree with you that trying to do a comparison based on small photographs of two completely dissimilar images is not a very good comparison.

But "less-than-1080p computer screens" ? If it were not for the fact that they're too small, I'd be using one of those 30" 2560x1600 monitors that are gaining popularity these days. (double the resolution of a 1080p display)

Even with 4K, you can only go to 44" in size before resolution drops below those 30" monitors. (100 pixels per inch)

If you want something suitable as an HDTV and a monitor, you really need 8K to have a large display with 100+ PPI resolution. Even then, computer displays are going to start exceeding 200PPI by the end of the year, which brings that size back down to 44" to have comparable image quality to a computer display.

You say we don't have to resample and then finish by saying we are upsampling . It is not like the original pixels are you still using have the same size. They do not. For the same screen size, they have shrunk and hence perceptually, will create a softer image as I showed in my post. Fact that you are using 2X resampling simply means you are controlling the accuracy errors. Otherwise, you will have some distortion created by that too.

If we merely doubled each existing pixel, we'd still be technically "upsampling" and yet would have an image result identical to the native 1920 x 1080 capture. Would that be "softer"? No, it would be identical. yet it's still technically upsampling... in this case using an algorithm programmed to maximize one very strict definition of accuracy. With subjective human visual perception there are other algorithms that may yield a subjectively better result than mere doubling despite shrinking the relative size of each of the original pixels in relationship to the newly interpolated ones. Hasn't this already been debated and demonstrated many times in this forum?

In an analog paradigm, any change to a signal can be described as "noise" that deviates from the goal of fidelity. But digital algorithms aren't bound by the laws of analog noise... they can think and do things that work with how human vision and perception work. If it's technically distortion because it introduces change, it can be helpful distortion rather than hurtful distortion: this is what a good digital algorithm should be designed to achieve (yes, I'd prefer no upsampling to bad upsampling). The smoothing of square-shaped edges to create visually smooth and continuous contours, as were represented by the real image prior to quantization, is an example. That's why you prefer your DVDs upsampled as you say. Why not 1080 images upsampled as well in that case? Are 1080 images really able to capture detail beyond the limits of human eyesight?

Even if someone felt that the upscaling improvement was negligible from a 30 degree angle (and not all would), anyone hoping to enjoy films from less than 1.5 screen widths would benefit from good quality 4K upsampling without question.

Technically since the four pixels have a gap between them it is not the same as the original single but we can let that go .


How is that in support of objecting to my use of the term resampling?


I sure hope not as that would void the basic mathematics and science of sampling . You can distort the frequency response as you resmple and with it, create a perception of sharper image but that has nothing to do with resampling.

The *ideal* resampler gives you the same image with finer pixels. But zero additional detail. Again, this is the ideal resampler. The non-ideal resampler has different frequency response and will either ring in the process of sharpening the image, or soften the image.

It is a myth that by resampling you create any new detail or resolution. Look at it in reverse. If I have a 4K image and create 1080p out of it, sampling theory mandates that I chop off anything with higher resolution than 1080p. This is a one-way function and there is no linear transformation that gets you back the information that was thrown out.

Here is Sony's white paper on their 4K projector:

"Next, every Sony 4K projector can use Intelligent Smoothing, a Sony circuit that scales an incoming HD or 2K picture up to the resolution of the 4K screen. In the process, Intelligent Smoothing interpolates new display pixels "in between" the source pixels. In HD-to-4K upscaling, the projector circuitry creates three new pixel values for each pixel in the source signal. These new pixel values do not represent additional picture information. (That can only come from the source.) But the interpolated pixels do enable a more seamless display, for a visible reduction in stairstep jaggies."

You can opt to sharpen the image in the process of interpolating the 1080p video back to 4K but that is an artificial manipulation of the image that creates artifacts. Here is an enlargement of what folks just post on the Sony 4K which I have am showing with or without Reality Creation turned on which accomplishes this:



As you see when we do try to sharpen, we create those halos and accentuation of compression artifacts.

Trust me guys. There is no magic here. I wish there was. We could make blu-rays out of DVDs then . But there isn't. The mathematics don't allow it.


Sorry don't know what the rest of this paragraph is other than the last sentence . You like your DVDs upsampled because relative to the size of our screens the pixels would be too large and very visible otherwise. Once we go to 1080p, for you typical seating position people essentially don't see the pixels anymore. Once that gets accomplished, then going to finer pixels does nothing.


Because all things are relative. With the first 6X in pixel increase we accomplished a lot. Once you get to your target, then additional resolution does do us any good.

You talked about perceptual effects. That is why I said upsampling actually makes the image softer. It does that because larger pixels tend to accentuate image sharpness. Smaller pixels do less of this and hence, serve to make the image softer. This is not related to resampling but the nature of how the pixels are presented. Remember: your video was mastered on a 1080p display. It was sharpened for that. If you go to a finer geometry, then you are seeing a softer image than the talent approved.


Again read the Sony phrase above. If you need more, read my posts in this thread: http://www.avsforum.com/avs-vb/showt...359018&page=40

I have gone over all of these points from many angles. No one is left arguing with me anymore . Simply put, resampling only has one purpose which is to reduce pixel sizes. If you are not able to see the pixels, you don't have a problem that needs fixing with 4k.

resampling is just to reduce pixel size is like saying MCFI is just to increase fps. There is perceivable logic for MCFI in LCDs.

Reducing pixel size reduces jaggies and can improve gradations when upsampling and approximate analog smoothness. That's the logic behind calculus and of course up to a perceivable point. They are all new material but it is not native. There is no new info (if that's what it is meant by "detail").

I don't understand why in this thread we keep making strawman out of this simple logic:
VCD< DVD< DVD upsampling< Blu Ray < Blu Ray up sampling < 4k
Why is the argument always on Blu Ray upsampling will not be as good as 4k native? Is that a revelation?

What I don't get is how bigger pixel can be sharper. I don't see that on my old VGA LCD monitor vs 1080p monitor. And if we go into pixel structure then we might as well talk about how different Sharp Quatron is with Samsung's PVA or with pentile. Like I posted earlier, if pixel is bothering us, then something is wrong somewhere. We do not see pixel. We see a blended image which pixel is one input. Neither do we see dpi on print but on aggregate there is an impact whether you using 300dpi or 1200 dpi

But I do get is that the perception of sharper may not even be a function of pixel but a function of enhanced contrast, which is one of the algo in Sharp's ICC. The eye is first and foremost more sensitive to contrast, hence the infamous Edge Enhancement technique in mastering.

Let's assume that's true for the sake of your argument.

If it was, then still, why would 1080 not benefit from upsampling? If I pull up a 1920 x 1080 desktop image from my PC I can see the limits of the 1920 x 1080 pixels from between 1 and 1.5 screen widths without having 20 20 vision. Are all HT viewers mandated to sit father away than 1.5 screen widths by a law of cinema? 30 degrees is recommended as a minimum viewing angle, not a maximum for peripheral stimulation, so why not enjoy it as a minimum rather than maximum in our HT venue as well? Why not a HT system that's friendly to front-seat viewers as well as mid-to-rear seat viewers?

Being able to enjoy native 1080p-encoded images (via 4K upsampling) without visible pixel structure from between 1 and 1.5 screen widths but maintaining clarity would benefit front-row viewers in many viewing rooms, and create for a more immersive image for those wishing to view constant width 2.35:1 systems which attempt to widen, rather than shorten, scope pictures.

Happy with that?


Agreed.

Bjoern's exellent examples of the 5th element upconverted DVD demonstrate how upsampling can better reveal detail that's inherent in the native image but obscured by quantization/pixel noise. Such algorithms interpolate new pixels, but not additional detail. I'm suggesting that 1080p images also have quantization noise and can have their native detail better delivered by careful upsampling as well. No "new" detail need be invented... just good algorithms to reduce quantization noise to best reveal the source's inherent detail.

If 1080p captured all possible visible detail/showed no quantization noise at 1.5 screen widths, then native 4K at the 1.5 screen widths would look identical to native 1080p without any detectable difference whatsoever. If 4K looks better at all (all other variables being the same), then this demonstrates that quantization noise is present and visible at the 1080 level (hence, the potential for 1080p to benefit from upscaling to reduce quantization/pixel noise and improve visible detail as a result).


Sounds good to me. Apparently Sony feels that this reduction in stairstep jaggies benefits native 1080p images. I don't disagree.

Damned by faint praise .


I explained that David.


I have not said any of this to you. I have said the following:

1. Upsamping results in no new data, or increased resolution/detail. None. The math dictates this so it is not an opinion.

2. If from your seating position you see the pixel structure, then getting a projector with finer pixels will solve that problem.

3. Due to realities of pixel structure, when you resample to finer pixels, subjectively the image will look somewhat softer. Not a lot. But somewhat softer. People like to talk about these images being "smoother." The way it become smoother is to have less sharp edges.


Vast majority of people are not seeing pixel structure from 1.5X. Not even close. If you are sitting closer than that, then likely you are seeing convergence errors in your projector, artifacts in the image, etc. Getting finer pixels doesn't solve these problems.


Context is important. As I said, no one is saying that you should watch 480p video on a 10 foot screen. You will see the pixel structure and therefore a 1080p projector will show a much more pleasing image.

There is however a crossover point at which the pixel structure is just adding enough high frequency noise as to make the lower resolution image look sharper. Since you are a believer in Buoern's work, here is an interchange documenting that from the archive: http://archive.avsforum.com/avs-vb/s...?postid=691456

Poster: but even then, the hard edges of the pixels add artificial high frequency content to the image. That's why we like to use higher resolution modes, and scale up to them.

Bjoern: Thats right. The higher you scale, the less aliasing noise (hard edges) you get.

Again, if the pixels are quite large and visible then the harshness becomes a liability and resampling helps a lot with that. But in tiny amounts which is typical in our viewing distance with 1080p, it can actually be beneficial. And getting rid of it will make the image look a bit softer subjectively.


Hopefully you see from above that as non-intuitive as it might be, the quantization noise can be your friend. It wasn't in 480p world because its level was huge. Not so with 1080p.

And again remember that your content has been sharpened relative to a 1080p display in post. Not 4K. If you get rid of that extra quantization noise, you will have a softer image than the talent approved.


There are no fast and lose formulas like you are using. A typical PJ at less than $10K will have huge amount of lateral CA, lens issues, non-flat MTF, AT screen, etc. that all contribute as different types of filters relative to you seeing the pixel edges. At what point you see the pixel structure is highly dependent on these formulas which in all cases, help hide pixels more, than not. I know with our $8K JVC if I got close enough to see the pixels, I would see a rainbow of colors due to CA.

Here is a shot I have from a $65K three-chip DLP projector:


You can see more than a pixel bleed. So there are no quantization noise as you can imagine from simulations people post on forums. Here is another shot showing the offset of each color:



As you can clearly observe, the thee primaries do not line up. As such, pixels do not just go from white to black but white to other shades and then black.


Who conceded that 4K looks better? At typical viewing distance you may not see the difference at all.

At CES Panasonic had a 4K display but in a small size. I want to say 24 inches. They had the exact same size display showing 1080p. I stood there for five minutes and only occasionally I could tell the sharpness difference. This was from ~2 foot distance. If I stepped back at all the difference would easily vanish and at any rate, on many scenes it was not different. Both displays looked great and folks would have been happy with either.


So you have to determine first if from your seating position you see said jaggies which I can tell you, is very hard on real video material instead of computer source. And further, even if it is there and you remove it you are going to get a softer image per above.

Please guys. Don't believe in video alchemy. There is a world out there that wants you dump huge amount of money toward 4K. But they couldn't even be bothered to give the source in that resolution. Don't fall for it. Know that it only solves one problem (pixel structure being too visible) and it comes at a cost both in softening the video and of course in the display, AVR, etc.

I think it's sharper if you're feeding two displays the same signal which is native to the lower resolution display. I have two projectors, one 1080p and one 720p which project on the same 90" screen. I sit close, 7 feet from the screen. The differences are pretty obvious when viewing the same 720p material on the different projectors. One particularly good example I've seen is the loading icon in a recent PS3 game which is two perfectly crossed lines in a square. On the 720p projector, it's clearly stair steps. On the 1080p projector, it's much smoother and only faintly looks like stair steps. This is nice, and all things being equal I prefer to view 720p content on my 1080p projector. But, no doubt, the 720p projector is sharper with 720p content. On the 1080p projector it's smoother, softer. I prefer this softer image because the pixel structure is less visible, but I certainly don't think it's as sharp as it is on the 720p projector.

People often say a 720p projector is actually better for console gaming than a 1080p one because of this difference in sharpness. I don't agree it's better, but I do think there's merit to it. It's a trade off. I think you could probably say the same about watching 1080p content on a 1080p display versus a 4k one. Sharper vs softer with less visible pixel structure.

Perhaps it's not a valid comparison though? Perhaps 720p can't scale up to 1080p as well as 1080p scales up to 4k?

Native to native is always the best.

Point is if the same analogue image/ video is mastered in VGA and in 1080p, how can VGA be "sharper"?

And yes softer image can actually be good for motion perception. Reason why calibrators tune sharpness in TV down.

No argument here then, native 1080p looks much sharper on my 1080p projector than on my 720p one and I think native 4k will look perfectly sharp on a 4k display.

However, that just doesn't seem to be the case with upscaling. While my 720p projector has larger pixels, it appears sharper than feeding the same 720p image to my 1080p projector. It makes sense to me that upscaled 4k will be likewise be softer.

Not anymore I am telling this but else: With the cognitive PQ improvement engines the 4K pics will look out of this world even from a distance .


Bravo&cheers, full agreement here

BUT, don't mix the front-row cinema viewing scenario with the TV viewing scenario as ppl understand it. Tell ppl truth that 4K is for the front-row lovers and that casual viewers in their family rooms will see nil difference. The catch is of course that by telling truth you can't hope to sell 4K TVs except to the microscopic niche of front-seaters. So the truth has to be hidden and 4K is PR-ed as the next gen 'TV' and not a 'front-row TV' .

On my Sony LCd's there was always the question 'do or do i not use the Reality Creation option when watching SD (on the 1080p TV)'.
Well, they should call it Artificial Digital Creation because that is how it looks and that is what it does, it ads a fakeish look to the picture.

TV's we'll buy are minimum 50 inch, twice as big as a 24 inch. Since a lot of stuff on the 4K TVs will be upscaled one of the big questions in this thread is 'how will upscaled 1080p look on a 50+inch 4K TV'.

A 50 inch set is over 300% larger than a 24 inch.
http://www.cavecreations.com/tv2.cgi

But I get your point. I'm looking for something in the 80 or even 90 inch range, for myself. I have been spoiled rotten with my 70 and 80 inch Sharps. I just can't see myself buying a smaller TV ever again. With my veiwng distance at about 10ft, I am hoping to see some benefit out of 4k.

But, I am taking a wait and see approach. As Rogo pointed out early on, most of what he saw was static or slow moving images. I want to see how it looks in the real world.

Only if you are sitting at such a distance from your display that you cannot see aliasing in the source.

If you can see aliasing in the source from the distance you sit (and I do, with 1080p) then upscaling can look better. Better is not necessarily sharper though, in all likelihood, it will be less sharp, unless you are going to sharpen the image after scaling. With selective/adaptive sharpening algorithms, this can actually be a viable option to increase the appearance of detail without adding visible ringing.


Who has been saying anything else!?

So you're granting that it's possible that something like the Sharp tech -- even if maybe it doesn't need the associated 4k, but rather only the iCubed part -- might actually look better to us humans when sold to us as a 4k display?

I wonder what Amir has to say about that. I'm inclined to agree with him on the relative pointlessness of Panasonic's 4k demo (although it was clearly superior, you had to get up close and very personal); but I'm 6 weeks out and still can see the Sharp's superiority over the 1080 in my mind's eye.

The only point which is debatable is if the 4K is a must for cognitive algorithms. I can accept it is a must due to some peculiarities but I doubt it. Another issue is that, at least with present algorithms, they are not infallible. Meaning that occasionally processing artefacts are showing up. This leads to a situation common today which is that there are tons of various picture processing functions in TVs but general advice is to turn them off for highest PQ. Obviously it is not excluded that the Icubed processing is so refined it is always better than the non-Icubed 4K. Then, the merits of 4K would be rather attributable to the Icubed than to the 4K per se.



Logic tells this:

1. If there is aliasing showing up in the source it means the source was not properly bandlimited and this should not be the case.

2. Upscaling is not solving aliasing problem. One can have even more aliasing by upscaling. Upscaling combined with proper bandlimiting will of course eliminate aliasing but then the point is why not to apply just bandlimiting without upscaling?

That can be true, but it is also an indicator that there is more resolution in the source material than 1080p is capable of.

Perhaps aliasing is not the correct term, but what I was actually meaning was you get close enough to a display, you start seeing "stair-step" artefacts with anything due to a lack of resolution.

I have to get about 6ft away from my 46" screen before this stops being a problem. (over three screen heights) At this point, the TV only fills a tiny portion of my view (30°) and I can't enjoy films/gaming on it. Normally I have a 50-60° FOV depending on whether I'm leaning back or not.

I expect this is fairly common for anyone using their "TV" as a multipurpose display, or anyone who has or has previously used a projector.

Good upscaling will reduce/eliminate aliasing.

Bandlimiting further reduces resolution below 1080p.

Aliasing on 1080p material comes mostly from the source which is either the digital camera or a badly done 2K scan of film. Bad compression or just bad work by the guy who does the authoring (DVD/BD), which happens frequently, just makes the aliasing in the original source worse.

But clean 1080p source does almost not exist, except from 4K or higher scans.

The one on the left is a simulation of native on an LCD TV, and the one on the right is a simulation of the same source on an LCD TV of the same size, with 4x the number of pixels (no special scaling - each source pixel just duplicated). I think the 2nd one is probably more accurate.