This race for more pixels is misguided. The change I most want to see is deeper pixels, at least 10bpp, preferably more. I'm getting really tired of all the banding I see on what should be smooth gradient images. If a landscape aspect image is P pixels wide then to display a gray scale gradient, black to white, you need
ceil(log2(P))
bits per pixel. So a "2K" display needs 11 bits, "4K" needs 12, etc. Then each column of pixels gets a distinct value and there is no banding.
So sure, give me a 4K screen. I've seen them and they are sweet, but you MUST also increase the pixel depth at the same time, or there will be artifacts.
And don't get me started on the horrid compression artifacts on basic cable. A crime against quality imagery.
Dithering can help a lot when it comes to displaying large smooth gradients in lower bit depths. Obviously it has its downsides, though - for one, you need to actually implement it where smooth gradients are used, and at when it comes to video games this basically never happens and you end up with notable banding instead.
Dithering in the field of video is pretty common, though. But it has a pretty large problem there as well - since dithering is essentially noise, it requires a lot of bitrate to compress efficiently, and if you don't have bitrate to throw at your source, you're most likely going to kill it and just end up introducing banding. Blu-ray is pretty much the main avenue where you have enough bitrate to spare for proper dithering in 8-bit video. Anything less than that, though... well, let's just say that House of Cards on Netflix was suffering from banding a lot.
Banding is actually one of the biggest reasons why anime fansubs these days are generally encoded in 10-bit H.264. Anime tends to have a lot of large and smooth color surfaces and banding was pretty much the hardest thing to avoid with regular 8-bit video - 10-bit on the other hands makes it an almost total non-issue. And for non-10bit displays, it moves the necessary dithering to the playback end, which is obviously a much nicer alternative since you don't have to compress any of that in the video itself. And beyond the gradients, 10-bit H.264 actually gives you better compression quality in general, which just makes it even better.
Now, it obviously comes with the downside of not being supported by hardware decoders anywhere, so you basically will need a decent CPU to decode 10-bit video. For fansubs and the people who make them this isn't that much of an issue though, since the advanced subtitles they use are also generally poorly-supported by hardware players, and this has been the case for a long time.
Next-generation video formats may actually bring higher bit depths to hardware decoders as well, though - H.265/HEVC has a Main 10 Profile intended for consumer applications.
This assumes infinite color discrimination abilities, though. I'm fairly certain that after a certain point, even if there are bands, most people would be unable to perceive them.
8 bits isn't "current", if you want "current" you get 10 bit panels. If you want "better than bargain basement" you get 10 bit panels.
At least where I live one of the differences between the TVs that CostCo sells and the ones that others sell are 8 bit (CostCo) vs 10 bit (others) panels.
I understand about the panels, but what about the source material? Is color information on a blu-ray stored at 8 bits per channel or is it higher? If the source material is 8 bits per color channel, I still think you'll see banding on smooth gradients no matter how good the panel is. I should try some tests...
Note that the panels are very unlikely to be true 10-bit; they're almost certainly 8-bit with temporal dithering. And the cheap panels are 6-bit with temporal dithering to 8-bit.
That is a good question. And strangely not as straight forward as I might like. "RGB" is 8 bit but video is not encoded with RGB, it is encoded YbCbCr [1], and the conversion from Chroma space to RGB space is fractional and depends on the Kr and Kb constants.
The math supports converting a YbCbCr 'pixel' into a 30 bit RGB pixel if your conversion constants allow for it. What is unclear is when quantization noise from the compression of the chroma signal shows up as bit instability in the resulting conversion cycle.
For TV I have no idea, but for a computer, you can get 10 bit per channel output from most graphics cards these days. You still have the problem of the source material being 8bit per channel though.
The banding you see is from image and video compression, not a lack of bpp. Human beings can almost never differentiate 6 bpp from 8 bpp grayscale, and with colors 7 bpp/c is almost always sufficient.
This is true, but transmitting video at 8 bits per channel is problematic because it's too close to the threshold of perception. Any video processing you do on an 8 bpc signal is likely to introduce visible artifacts, and of course today's TVs go crazy with ridiculous video processing even at their default settings before people start playing with the brightness and contrast controls. We need to be transmitting video at 12 bpc so that video processing done at the display doesn't destroy the signal.
I tend to agree, but I would prefer to discourage television manufacturers from doing any more "improving" than they are already doing. If you're sending a signal to a display with the signal so little compressed that 12 bpc would be an improvement for post processing, then you're probably sending the signal straight from a BlueRay player or desktop graphics card. In both of those cases, I'd rather my TV not do any post processing. (Brightness and contrast on TVs are analog operations and do change with changes in bit depth. Higher brightness = turn the lights up, not a pointwise linear map)
This has been in the HDMI spec for 7 years already and most high-end TVs support 10-12 bit input. And lower end TVs it doesn't matter anyway. Also TVs often use at least 10 bits between internal processing steps.
Anyway, what you want is happening for 4k - Rec. 2020 mandates at least 10bit for 4k
Reading a few pages in, if you're not seeing it on the market, it's because the technology is patented (and it's the obvious approach of using an array of LED backlights with DSP correction for the optical scattering of the backlight filter).
As soon as there's the slightest amount of noise, it's impossible to discern between 8bpp vs 10bpp. And real-life applications will almost always have noise.
It's probably going to be like the processor wars. It'll keep going up until they reach a peak and then it'll switch to multi-cores (or in this case, expanding the number of bits).
32-bit is 8 bits per colour channel (24-bit) plus an 8-bit alpha (transparency) channel, not 10-and-a-bit bits per channel. The alpha isn't (normally) sent to the display device, it's used to render overlapping elements at the display generation level (video card). Good LCDs can easily render the full colour depth with headroom to spare. Higher-end graphics monitors are often 12-bit (the extra range is used for calibration/display accuracy, not necessarily for greater depth).
ceil(log2(P))
bits per pixel. So a "2K" display needs 11 bits, "4K" needs 12, etc. Then each column of pixels gets a distinct value and there is no banding.
So sure, give me a 4K screen. I've seen them and they are sweet, but you MUST also increase the pixel depth at the same time, or there will be artifacts.
And don't get me started on the horrid compression artifacts on basic cable. A crime against quality imagery.