Screensize & resolution & distance

[Phil]

Yeah, I agree.

I actually fiddled around to get a practical example - I took a 1920x1080 screenshot of KZ:SF and croped a quarter of it (960x540). I then enlarged it to 1920x1080, once using a pixel doubler (every pixel becomes double size) to pose as our native resolution, and second, using bicubic which will pose as our interpolated image.

The first one is technically our 'native' picture of 960x540 and the second one, the interpolated one.

960x540 to 1920x1080 is the same increase as going from 1920x1080 to 3840x2160 (4k).

By opening them in two browser windows, you can switch back and forth between them to see the difference. When viewing them up close on your computer screen, just remember that you should technically double your viewing distance as on the doubled shot, the pixels are double in size (one pixel becomes 4) to have a representative image of a 1920x1080 screen viewed on a 1920x1080 set.

http://temp.conceptics.ch/b3d/kzsf_doubled.jpg
http://temp.conceptics.ch/b3d/kzsf_interpolated.jpg

I think that should be fairly representative. :smile:

 

[Deleted member 11852]

Not sure how well it works in practice - but it's a brilliant idea nonetheless.

I'd mentally tagged X-Reality Pro as 'snake oil' technology but having put both sets side by side, the 4K set produces more pleasing (subjective) upscaled SD content than the old 1080p set. This is contrary to what I expected from video sources with so little little information to work with and where I expected a larger screen to amplify video artefacts.

 

[brunogm]

@RudeSoup if you think the Modulation Transfer Function is a lie, so is Fourier Transform and Nyquist and the possibility of display anything with a finite number of points.
I'm ok if gamers "see" more than average people maybe 20/10 in zed amazing site. THX and the ImageScienceFoundation determined the "rule of thumb" 1.5H to obtain a good FOV generally. Also distance changes a meter(3feet) between relaxed and excited "edge of the sofa" states. The 900p, 810p, 1080p battle is for efficiency of resources to achieve some goal. What is your goal?

 

[sebbbi]

Games have way more high contrast sharp edges than movies. This is because games don't have perfect antialiasing, every frame in a game is a snapshot with zero exposure time (1/48s accumulation of light of a single movie frame vs perfectly sharp image with no motion) and there's no depth blurring outside the focal distance (infinite depth of field).

It's very hard to make a game that uses camera depth of field in a movie-like manner and is still playable. In movies there's no interactivity, the director tells the story with frequent camera cuts (different angles of the same scene, close ups) and focus chances (blurring the area where you are not supposed to look). The camera focus changes follow the actions (and the sounds) on the story closely. This makes it easy to follow the focus changes with your eyes. For example when there's a discussion between two characters, the focus will shift to show the face of the other guy just before he starts speaking. This is how we naturally follow a discussion. If the focusing is directed wrong, the scene feels awkward.

The game however is fully interactive, and it's very hard to estimate where the player wants to look in the screen. If the game focuses on the wrong area of the screen and you look at the blurred area, it feels like you have problems in your eye-sight and you get a headache. We need eye-tracking before we get good camera depth of field simulation in games (outside cinematic sequences of course).

Motion blur has similar problems in games. Motion blur should be also based on eye movement. In movies the director controls your eye movement, so the motion blur works much better compared to games. Also the camera is static (mounted on a tripod) in most scenes in movies. Camera movement is often handled by cuts. Of course there are scenes with panning (or camera in rails), but even there the director knows roughly where you are looking at (or has the option to use a small aperture to make it all sharp). In games the camera (both target and camera itself) are constantly moving based on player inputs.

Antialiasing is at least a solvable problem with current generation hardware. Some combination of 4xMSAA and post process AA should be enough to hide most of the overly sharp edges. But without proper lens, focus and exposure simulation the images are still unnaturally sharp (compared to movies and/or our eye sight). Games also tend to have small (and very sharp) text (UI) compared to movies. Because of all of this, games need displays with higher resolution than movies to look perfect.

 

[Shortbread]

1080P content on a 4K screen is going to look worse not better, I think

My Sony/Samsung 4K sets disagrees... actually the additional scaling provides a cleaner look and response with 1080p material, than my 1080p sets.

 

[Globalisateur]

My Sony/Samsung 4K sets disagrees... actually the additional scaling provides a cleaner look and response with 1080p material, than my 1080p sets.

I agree (in theory).

I don't understand all this talk about "upscaling with blurring" when talking about 1080p -> 4K. The scaling should be perfect, one "1080p" pixel should directly become 4 identical pixels at 4K.

There shouldn't have any upscaling artifacts or loss of information, right? Except on some TVs maybe who don't take the whole native 1080p to be properly scaled to 4K?

EDIT: Ok now I realised that maybe some 4K TVs do some bilinear interpolation upscaling even from 1080P to 4K instead of nearest-neighbor interpolation upscaling? Well, better avoid those 4K TVs then!

 

[Shortbread]

EDIT: Ok now I realised that maybe some 4K TVs do some bilinear interpolation upscaling even from 1080P to 4K instead of nearest-neighbor interpolation upscaling? Well, better avoid those 4K TVs then!

I agree that all TVs aren't made equal... and the more expensive ones tend to have better interpolation hardware. I would imagine a $600 55' inch 4K set isn't going to compete very well with very expensive ones when it comes to proper IQ scaling/interpolation. Example; most I080p material I've seen on Seiki 4K sets, seems soft and grainy in areas. My Sam/Sony offer great scaling and response, but I also paid a pretty penny for those models as well.

 

[Shifty Geezer]

EDIT: Ok now I realised that maybe some 4K TVs do some bilinear interpolation upscaling even from 1080P to 4K instead of nearest-neighbor interpolation upscaling? Well, better avoid those 4K TVs then!

As I mention above, if you have the same FOV, even a bilinear interpolation won't look any worse. If you're sitting close enough that a 2x2 grid of 4k pixels looks blurry, 1) That same blurred group would look like a discrete square at 1080p which is hardly better, and 2) You're too damned close!

I was watching some Mythbusters on my 1366x768 TV. SD content is upscaled very well by and large with no sign up scaling until you hit some graphics. Then straight lines become very pixelated with a lack of temporal differences to build the frame from. That's what 1080p > 4k will be like. If the upscaling fails, at worst you'll have some 1080p quality.

 

[Deleted member 11852]

@RudeSoup if you think the Modulation Transfer Function is a lie, so is Fourier Transform and Nyquist and the possibility of display anything with a finite number of points.

Nobody, me included, is denying these approaches for scaling. What I was skeptical about, having seen Sony's X-Reality Pro PR, was that the technology would live up to the promises. That's really for each individual to decide but I am nonetheless very impressed with the scaling. The bigger you go (I went from 40" to 49") the more poor scaling from low resolution source material will be evident but the TV does a great job.

The 900p, 810p, 1080p battle is for efficiency of resources to achieve some goal. What is your goal?

I didn't have a goal. I was browsing the forum last week and Graham posted about the TV he had then I simply had to buy a new TV. I'm not quite sure what happened to be honest, I kind of blacked out. That was the explanation I gave my girlfriend anyway

I was watching some Mythbusters on my 1366x768 TV. SD content is upscaled very well by and large with no sign up scaling until you hit some graphics. Then straight lines become very pixelated with a lack of temporal differences to build the frame from. That's what 1080p > 4k will be like. If the upscaling fails, at worst you'll have some 1080p quality.

Yeah, it's very evident on The Daily Show too. Everything looks good until your eyes fleeting cross the Comedy Central Logo in the corner.

 

[Shifty Geezer]

I'm not quite sure what happened to be honest, I kind of blacked out. That was the explanation I gave my girlfriend anyway.

Shopping Induced Epileptic Tendencies. It's a real problem that needs proper scientific investigation, and honest, understanding support for the victims.

 

[Deleted member 11852]

Some kind of telethon is probably in order!

 

[brunogm]

SchubinCafe delivers again!!!

What’s hype, what’s real, and what’s just not known yet? Join Mark Schubin in this recorded webinar as he looks at 4K (UHD), high frame rate, high dynamic range, and wide color gamut to see which really gives viewers the best picture, and what it will cost you in resources to provide. The best picture may not be what you think it is, but may be where your organization is heading. /QUOTE]

[/ http://www.schubincafe.com/2014/...keting-is-not-engineering-recorded-webinar/ ]