[opendtv] Precision
- From: dan.grimes@xxxxxxxx
- To: opendtv@xxxxxxxxxxxxx
- Date: Thu, 24 May 2007 15:53:33 -0700
In video, how much precision is enough?
Let us take audio as an example. Audio can be sampled at 48,000 samples
per second, 16 bit, and it is "CD" quality. But sample at 192Khz at 24
bit, process, downconvert, encode, etc., and the quality is markedly better
than when produced at the previously stated format. Of course, we are only
dealing with one axis, representing pressure, and time (or are we?).
However, if we sample audio at 384Khz at 48 bits, I bet not but three
people in the world could tell a difference from the audio produced at
192Khz at 24 bits. So there is a point to which we have human limitations
and there is no reason to sample higher or with more "precision" (Law of
diminishing returns), save research areas.
However, there is a weak link to this aforementioned audio recording:
spatial offset. Even with 7.1 surround sound, there is still spatial gaps
to fill with audio. So there is still room for improvement. But it is not
in sampling frequency or bit depth that will make the difference.
Now back to video.
First of all, bit depth: 8 bit, 10 bit, 12 bit, higher? What is the
graduation difference the eye can see at a particular wavelength?
Obviously, it is going to be different at different wavelengths. So what
is our target maximum? Or should it be different for different
wavelengths?
I can clearly see a difference in each of the currently used bit depths;
all are still well below the human's ability to perceive. The contrast
range of the human eye is incredible.
I liked the papers that were recently presented about Color Gamut. Clearly
there is progress in developing better displays and current production
formats help deliver the quality for those displays.
Frame rates could improve. Temporally, we can perceive even changes at
1/60th of a second. When I did high speed photography, I was amazed at
what my eye could see based on the effective shutter speed I needed to
capture the image. Of course, to me it was a blur, but I knew it was
there. Maybe, like the color gamut, the question is not what can be seen
but what is the fastest object we are likely to capture on film, making
that the maximum. But I would suggest that it could be in the order of
1,000 frames per second or more.
And then, there is peripheral vision and emersion effect. While IMAX does
indeed cover a very large area, even in the optimal seating position, it is
not optimal. I always liked OmniMAX better (where did they go?) because it
covered more peripheral, although not all. And as soon as one looks to an
object to one side, one sees border. Ultimately, one would need 360
degrees spherical, except maybe where the couch sits.
I am really am asking these questions so feel free to let me know if you
know some of the target maximums. Like audio, I think there are some
maximums where we can quit, but I don't think the current technology is
there yet. I am just curious if anyone here knows what they are.
But one could still produce audio in 48Khz at 16 bit, process and encode,
and still enjoy the music (or discussion, etc.). Almost anyone with normal
ears can hear the difference but might not care, so long as the content is
keeping our interest.
The same is most likely true of our High Def. formats today. After
writing this, it does seem like there is a lot of room for improvement in
the capturing and display of visual media.
Perhaps this is not really appropriate to OpenDTV so please feel free to
disregard.
Other related posts:
