[opendtv] Re: TV Technology: 4K for Broadcast: Is it Worth the Expense?

  • From: Mark Schubin <tvmark@xxxxxxxxxxxxx>
  • To: Open DTV Forum <opendtv@xxxxxxxxxxxxx>
  • Date: Sun, 2 Jul 2017 08:58:25 -0400

On 7/2/2017 2:11 AM, Manfredi, Albert E wrote:

Visual acuity has to be taken into account. Rule of thumb, ~1 arcmin of angular separation starts to become noticeable, for humans with 20/20 vision (or less than great vision, but wearing appropriate spectacles).

We could argue whether absolute visual acuity is as you describe (called 30 cycles per degree or cpd) or higher or lower. And there are situations when absolute visual acuity might matter. Watching television is not one of them. For watching television, what's much more important is the psychovisual sensation of sharpness, which is based on both resolution AND contrast.

If you are not familiar with the human contrast-sensitivity function, these composite images should make its importance clear:
http://bit.ly/angry-neutral
They were created by Aude Oliva at MIT and Philippe G. Schyns at the University of Glasgow in 1997 and are copyright by them. They are clearly not high resolution. Combined, they are only 479 x 236 pixels. You should see an angry man on the left and a neutral woman on the right. Now walk away from your screen. At some point, they will change places, with the angry man on the right and the neutral woman on the left.

The images are simple JPEG. They are not animated GIF, and, even if they were, I have no way to take control of your camera, judge your distance from the screen, and trigger the animation at the correct moment. If you prefer, you may print the images; they work just as well on paper.

What's going on? The images with the angry man on the left are intended to be seen with their detail at about 6 cpd or higher; the ones with the angry man on the right are intended to be seen with their detail at about 2 cpd or lower. The human contrast-sensitivity function is so strong that, under those conditions, the "wrong" images completely disappear.

If you graph the modulation-transfer function (MTF) of a system (the contrast at any given resolution), you come up with a curve heading downward as resolution increases. According to the image-quality work of Otto Schade at RCA [Schade, Otto H., "Image Quality : a comparison of photographic and television systems," RCA Laboratories, 1975, republished in the SMPTE Journal, volume 96, number 6, June 1987], the psychovisualsensation of sharpness is proportional to the square of the area under the curve; according to Erich Heynacher at Zeiss [Heynacher, Erich, “Ein Bildgütemaß auf der Grundlage der Übertragungstheorie mit subjektiver Bewertungsskale” [Objective Image Quality Criteria, based on transformation theory with a subjective scale], Zeiss Mitteilungen, volume 3, number 1, 1963], it is the area, not the square. Either way, sharpness is based on contrast in addition to resolution.

At the NAB Show in April, NHK (the Japan Broadcasting Corporation), perhaps the world's greatest promoter of higher resolution, showed the MTF of some near-ideal 8K image sensors. Contrast was 100% at zero resolution and zero at 8K resolution, as would be expected. At 4K, it was about 50%, at HD, it was over 80%. Add a lens and real-world imagery, and the high-resolution MTF goes down even more. Adding resolution extends the "toe" of the MTF curve, which has very little area under it; adding contrast (i.e., HDR) raises the "shoulder" of the MTF curve, greatly increasing the area under it.

Sony took advantage of that low area under the MTF-curve toe when it introduced HDCAM, dropping horizontal luma resolution from 1920 to 1440; Panasonic went even farther with DVCPRO HD, dropping from 1920 to 1280. Today, with improved technology, both companies' HD recorders capture a full 1920. Compare the older recordings with the new, and the difference is perceptible -- but it's not obvious.

Similarly, CBS has a 4K viewing area set up in its Manhattan lab. There are three 65-inch TV screens, viewable simultaneously from either three times the picture height or 1.5. All get content shot in 4K. One screen gets an HD downconversion and displays it as HD. The next gets an HD downconversion and upconverts it to 4K for display. The third gets 4K and shows it as 4K. CBS has made the setup available to just about anyone who wants to see the results, so what I'm about to report is not just my opinion but that of everyone I know who has seen it, including people from other broadcasters.

At the 3H distance, it's possible to see the difference between the HD and the 4K displays, but it's not obvious. It's almost impossible to see any difference between the 4K displays. At 1.5H, it's easier to see the difference between the HD and the 4K, but it's still not strong; it's also possible to see the difference between the upconverted HD and the true 4K, but it's not obvious.

So, all things being equal, sure: let's have 4K. But all things are not equal, and the increased sharpness going from HD to 4K is not great. Perceptible, yes; great, no.

The increased sharpness going from SDR to HDR IS great.

TTFN,
Mark

Other related posts:

  1. Home
  2. opendtv
  3. Archive
  4. 07-2017