You should watch this - AVS Forum
Forum Jump: 
 
Thread Tools
post #1 of 18 Old 09-25-2013, 06:11 AM - Thread Starter
AVS Addicted Member
 
mark haflich's Avatar
 
Join Date: Dec 2000
Location: brookeville, maryland, usa
Posts: 19,195
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 293
I strongly recommend you go the AVS Radio Show Forum, go to the Joe Kane on 4K thread, and watch the video at post 115. You will learn much there particularly about increases in dynamic range and HDMI 2.0. After a brief introduction in German, it is all in English and there is much information there most of it I bet YOU didn't know.

Mark Haflich
markhaflich@yahoo.com
call me at: 240 876 2536
mark haflich is online now  
Sponsored Links
Advertisement
 
post #2 of 18 Old 09-25-2013, 06:51 AM
AVS Special Member
 
blee0120's Avatar
 
Join Date: Oct 2009
Location: Merillville, IN 46410
Posts: 3,516
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 9 Post(s)
Liked: 52
Good to listen and learn on 4K
blee0120 is online now  
post #3 of 18 Old 09-25-2013, 07:04 AM - Thread Starter
AVS Addicted Member
 
mark haflich's Avatar
 
Join Date: Dec 2000
Location: brookeville, maryland, usa
Posts: 19,195
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 293
Interesting on not viewing further away than 1.8 times the viewing height for 4K, the effects screens have in generating noise and the support for gray screens. I think JK would recommend his .6 JKP affinity for a screen only 2M wide as in the German lab at the end of the video, but mainly to keep the screen within illumination standards. Note they are using a Samsung SP-a800 or 900 which is only 1080p. A screen width of 2M would not be adequate to observe all that is going on as a result of 4K mastering although, and I have modified my previous position on this, some of the benefits of 4K source resolution will be apparent on smaller than 10ft wide screens.

Mark Haflich
markhaflich@yahoo.com
call me at: 240 876 2536
mark haflich is online now  
post #4 of 18 Old 09-25-2013, 10:36 AM
AVS Addicted Member
 
stanger89's Avatar
 
Join Date: Nov 2002
Location: Marion, IA
Posts: 17,289
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 112
I haven't made it all the way through, but the first thing that jumps out at me is his comments about there being a (I can't remember the exact term he used) "huge" destruction of image quality scaling from 4096 to 3840, and determining this based of the results of artificially generated test patterns.

While that is true, it is completely not applicable to real world content. Single pixel wide alternating black bars are an "impossible" signal, it's essentially a square wave, and square waves have infinite bandwidth. Even if you were to approximate it, in order to sample a signal of 4096 columns of alternating bars, you would have to sample it about 14336 times (square waves have significant bandwidth up to about 7x the fundamental frequency, so 7*2048cycles).

These alternating pixel patterns completely ignore/violate the foundations of sampling theory, that being that the sampled signal is band limited to 1/2 the sampling frequency.

What I'd like to see, for/from all those folks who claim resampling is bad, is to take a picture of a physical test pattern with, say on a high end DSLR, render it down to say 2048, and then scale that test pattern to 1920, and compare the two and see just how much difference in sharpness there is. There will be a difference, 1920 can't be quite as sharp at 2048 (same with 3840 vs 4096) since it's lower resolution, but it shouldn't be a dramatic difference.

It's the same theory as resampling audio from 48kHz down to 44.1kHz, you can do that just fine without any major problems assuming it's done correctly.

See what an anamorphoscopic lens can do, see movies the way they were meant to be seen
stanger89 is offline  
post #5 of 18 Old 09-25-2013, 10:40 AM - Thread Starter
AVS Addicted Member
 
mark haflich's Avatar
 
Join Date: Dec 2000
Location: brookeville, maryland, usa
Posts: 19,195
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 293
And even if it were true, which I agree it wouldn't be if done correctly, the industry could just err pan and scan to eliminate 3 1/2 percent of the 4096 wide frame.

Mark Haflich
markhaflich@yahoo.com
call me at: 240 876 2536
mark haflich is online now  
post #6 of 18 Old 09-25-2013, 01:36 PM
AVS Addicted Member
 
noah katz's Avatar
 
Join Date: Apr 1999
Location: Mountain View, CA USA
Posts: 20,285
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 35 Post(s)
Liked: 125
Quote:
Originally Posted by mark haflich View Post

You will learn much there particularly about increases in dynamic range...

What kind of dynamic range?

If it'sfrom bit depth, I don't see the need.

Noah
noah katz is offline  
post #7 of 18 Old 09-25-2013, 02:29 PM
AVS Addicted Member
 
stanger89's Avatar
 
Join Date: Nov 2002
Location: Marion, IA
Posts: 17,289
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 112
Increased bit depth would reduce banding.

See what an anamorphoscopic lens can do, see movies the way they were meant to be seen
stanger89 is offline  
post #8 of 18 Old 09-25-2013, 03:26 PM
AVS Addicted Member
 
noah katz's Avatar
 
Join Date: Apr 1999
Location: Mountain View, CA USA
Posts: 20,285
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 35 Post(s)
Liked: 125
Right, not a problem for me, so not a reason to be interested in UHD; my not-quite-perfect lasik job makes it so that I can't even fully resolve 1080P.

Noah
noah katz is offline  
post #9 of 18 Old 09-25-2013, 04:05 PM - Thread Starter
AVS Addicted Member
 
mark haflich's Avatar
 
Join Date: Dec 2000
Location: brookeville, maryland, usa
Posts: 19,195
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 293
This is not the response I would expect to get from an engineer of your caliber. Banding can mean other things and arise from other causes. But here we are talking about gray scale steps and the numerical limitation imposed by 8 bits. Of course any bit limitation has its own limitation on the number of steps but going to 10 bits increases the number of steps fourfold. I am sorry to learn about your eye problem. Perhaps some fixes can still be done. Here's hoping.

Mark Haflich
markhaflich@yahoo.com
call me at: 240 876 2536
mark haflich is online now  
post #10 of 18 Old 09-25-2013, 05:20 PM
AVS Addicted Member
 
noah katz's Avatar
 
Join Date: Apr 1999
Location: Mountain View, CA USA
Posts: 20,285
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 35 Post(s)
Liked: 125
Why, should I demand less banding when I don't see it now?

I went back but they said they got close enough that if they try again it may actually come out worse.

I test as having 30/20 or 20/30 (whichever one is worse than 20/20) but the odd thing is that with high contrast, like lit signs at night) I see two clear double images from my distance eye.

Noah
noah katz is offline  
post #11 of 18 Old 09-25-2013, 05:48 PM
AVS Special Member
 
coderguy's Avatar
 
Join Date: Dec 2006
Posts: 9,951
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 139
A lot of banding is from the source interacting with how a device translates the colors I think.

Even having dealt with lots of graphics over the years (some professionally, some not), I don't know all the causes of banding. My guess is it will improve re-sampling, streaming, compression efficiencies, and some CMS interaction.
To some that doesn't mean much, to some others it might. To some devices, it might not fix much, to others it might?

Without the source being purer in its own form (hence still being susceptible to banding), perhaps there is still a lower chance of banding using a higher bit-rate with a bad source simply because the chances of 2 close colors being overlapped from the lack of precision in CMS errors would be lessened.

That said, I doubt this is going to have much visible effect on reference level content, unless it allows them to change the codec used and reduce the lossy-ness of some type of enhanced bluray (well an updated version) encoding (haven't seen too many superior formats on my own PJ in an objective and comparative manner, so who am I to argue).


Quick and Easy Shelf Mount Method for both one projector or dual stacks

Web Calculator v023 & v025
- Quick Peak at the new upcoming calculator
**Current Projector Calculator** -- http://www.eliteprojectorcalculator.com

Coder's Top Projector Picks of 2012 --http://www.avsforum.com/avs-vb/showthread....

coderguy is offline  
post #12 of 18 Old 09-25-2013, 07:47 PM - Thread Starter
AVS Addicted Member
 
mark haflich's Avatar
 
Join Date: Dec 2000
Location: brookeville, maryland, usa
Posts: 19,195
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 293
Any non uniformity in a color field, gray field, what ever is banding. Also there only 255 gray scale steps available with an 8 bit signal.As you blow the picture up, you can see steps in say facial tones where the color should blend. This is banding because there are not enough graduations to fool your cleaver sharp discriminating eyes. As the resolution increases and the dynamic range increases, banding worsens.

Mark Haflich
markhaflich@yahoo.com
call me at: 240 876 2536
mark haflich is online now  
post #13 of 18 Old 09-25-2013, 07:57 PM
Advanced Member
 
rcohen's Avatar
 
Join Date: Feb 2005
Posts: 796
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 42 Post(s)
Liked: 47
I haven't watched the video yet, but...
8 bits is not quite enough for normal displays with smooth gradients, and you will see subtle banding.
Add 1/2 bit of noise before quantization, and the banding is imperceptible.
Often, cameras already have that much noise present, and some sources (including 3D rendering) add noise for this reason.
Unfortunately, that kind of subtle high frequency noise gets lost with lossy compression, although it works great for uncompressed signals.

Aside from compression, the other big culprit is multiple stages of quantization. For example, take an 8 bit source, process it (i.e. color correction), then send out an 8 bit signal over HDMI, then process it again, then show it on an 8 bit display device, you get 3 stages of quantization. This will result in strong banding. The solution to this is to do all the processing in higher precision, and never quantize back to 8 bits. For example, take an 8 bit source, process, transmit in 10 bits, process, then display on a 10 bit device.

Another culprit is high dynamic range. The better the dynamic range, the more bits you need to keep up with the precision of your eyes. 8 bits starts to fall really short as displays get better.
rcohen is online now  
post #14 of 18 Old 09-25-2013, 08:09 PM
AVS Special Member
 
coderguy's Avatar
 
Join Date: Dec 2006
Posts: 9,951
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 139
Quote:
Originally Posted by rcohen View Post

I haven't watched the video yet, but...
8 bits is not quite enough for normal displays with smooth gradients, and you will see subtle banding.
Add 1/2 bit of noise before quantization, and the banding is imperceptible.
Often, cameras already have that much noise present, and some sources (including 3D rendering) add noise for this reason.
Unfortunately, that kind of subtle high frequency noise gets lost with lossy compression, although it works great for uncompressed signals.

Aside from compression, the other big culprit is multiple stages of quantization. For example, take an 8 bit source, process it (i.e. color correction), then send out an 8 bit signal over HDMI, then process it again, then show it on an 8 bit display device, you get 3 stages of quantization. This will result in strong banding. The solution to this is to do all the processing in higher precision, and never quantize back to 8 bits. For example, take an 8 bit source, process, transmit in 10 bits, process, then display on a 10 bit device.

Another culprit is high dynamic range. The better the dynamic range, the more bits you need to keep up with the precision of your eyes. 8 bits starts to fall really short as displays get better.

Thanks for the detailed info, this is the most complete explanation I have seen thus far.
Quote:
Originally Posted by mark haflich View Post

Any non uniformity in a color field, gray field, what ever is banding. Also there only 255 gray scale steps available with an 8 bit signal.As you blow the picture up, you can see steps in say facial tones where the color should blend. This is banding because there are not enough graduations to fool your cleaver sharp discriminating eyes. As the resolution increases and the dynamic range increases, banding worsens.

That's true, but I don't usually notice banding artifacts on the better video devices of today unless streaming or watching older movies or poor transfers, but I suppose if I looked harder for them.


Quick and Easy Shelf Mount Method for both one projector or dual stacks

Web Calculator v023 & v025
- Quick Peak at the new upcoming calculator
**Current Projector Calculator** -- http://www.eliteprojectorcalculator.com

Coder's Top Projector Picks of 2012 --http://www.avsforum.com/avs-vb/showthread....

coderguy is offline  
post #15 of 18 Old 09-25-2013, 08:47 PM - Thread Starter
AVS Addicted Member
 
mark haflich's Avatar
 
Join Date: Dec 2000
Location: brookeville, maryland, usa
Posts: 19,195
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 293
Just stare until you start seeing 2D in 3D, then stop. smile.gif

Mark Haflich
markhaflich@yahoo.com
call me at: 240 876 2536
mark haflich is online now  
post #16 of 18 Old 09-26-2013, 12:56 AM
AVS Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 5,414
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 16 Post(s)
Liked: 108
Nothing new for me. Seems that even Joe doesn't know everything, though. Two complaints from my side:

(1) "720p is vastly superior to 1080i". Such nonsense, at least for movies. IVTC converts 1080i60 to perfect lossless 1080p24. You can even encode progressively to 24p and just flag the video stream to 60i. Some USA broadcasts and DVDs are doing that. It's called "soft-telecine". Actually that's exactly what HD DVD did.

(2) Upscaling of 1080p to 2160p "should be an easy conversion to do well". The usual linear scaling algorithms which have been used for decades (Bilinear, Bicubic, Lanczos, Sinc etc) do not care whether the scaling factor is a cardinal number of a fractional number. A scaling factor of exacty 2.0x does make it possible to use non-linear scaling algorithms (and both Sony and JVC are doing that), but those algorithms can be mathematically quite complex. For example, I believe Sony has a database of learned patterns ("Reality Creation") which they use to perform the upscaling. That's not an "easy" algorithm to apply. In the HTPC world we have NNEDI3 which is using a similar technique and it's quite slow and involves a lot of math. But it does give very good results.
madshi is online now  
post #17 of 18 Old 09-26-2013, 03:27 AM - Thread Starter
AVS Addicted Member
 
mark haflich's Avatar
 
Join Date: Dec 2000
Location: brookeville, maryland, usa
Posts: 19,195
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 293
Joe often is wrong. He doesn't understand every thing about video and has often made incorrect statement and drawn incorrect conclusions based on inadequate knowledge. But he also often is right and has done much good for the sport. His primary fault and he indeed has many, is that he feels he is intellectually superior and has far more knowledge than the audience he is addressing. He justifies using incorrect terms because he feels his audience wouldn't understand the use of the correct term. If a question or statement contradicts his position he tends to ignore it and answer some other question. But he would absolutely suck if he tried to be a politician. But we should be glad we have him.

Mark Haflich
markhaflich@yahoo.com
call me at: 240 876 2536
mark haflich is online now  
post #18 of 18 Old 09-26-2013, 07:11 AM
Advanced Member
 
rcohen's Avatar
 
Join Date: Feb 2005
Posts: 796
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 42 Post(s)
Liked: 47
GPU acceleration has opened the door to doing much more sophisticated scaling algorithms in real time. One example of sophisticated algorithms are ones that analyze the image, identify the angles of edges, and maintain a perfectly sharp, anti-aliased edge in that direction. Old-style fixed filter kernels are non-adaptive to the image, and they always apply the same kernel in all directions. Non-integer scaling can be done whether using fixed or variable kernels. You just have to recompute the kernel, or use filtered sampling from a precomputed kernel. The big change with a non-aligned filter is that you lose the ability to do separable filtering on each axis. With GPUs, all these things are possible.
rcohen is online now  
Reply Digital Hi-End Projectors - $3,000+ USD MSRP

Tags
Samsung Sp A800b Dlp Projector 1080p
Gear in this thread

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off