[opendtv] Re: State of the Art LCD 45" HDTV

At 9:15 PM -0400 10/28/04, Tom Barry wrote:
>It is nice they can present non-Nyquist images.  But I'm worrying that,
>to the extent you can at all see the pixels, they can't really do
>anything else.  And that may interfere a bit with whatever sort of
>mental interpolation our brains do with eye motion tracking.

If you can see the pixels, you are too close to the display. You 
could be critical of watching video on a computer screen for this 
exact reason, but for a dedicated video display should never be 
viewed at a distance where you can see the samples in the raster. 
This is analogous to sitting in the front row of the theater.

A properly designed video display system will use screen samples that 
cannot be resolved at the designed viewing distance. You can move 
inside this designed viewing distance at your own risk. And when we 
use a computer display to view video full screen, we typically do not 
sit as close as when we are browsing the web or doing e-mail. Back 
off a few feet and the raster (pixels )will disappear.

>
>With a nice fuzzy CRT spot beam if your image moves 1/2 pixel then
>adjacent pixels will tend to blend to grey on a sharp edge, with a
>gradient between them.  But on a fixed pixel display there is more apt
>to be a sudden change of value.  Any filtering can only change the value
>of the entire tiny little square pixel and if that pixel is individually
>perceptible it may cause a visible dithering effect on some motion.

This is simply not accurate. If the image is sampled properly the 
samples will have the characteristics you describe and a fixed pixel 
display will reproduce them accurately. yes, the CRT display may have 
a softer transition between pixel values, but this is an artifact, 
not a useful feature. The CRT spot is not reproducing the samples 
faithfully - it is bluring them - and it does this in a nonlinear 
fashion, with less sharpness at the edges of the screen than in the 
center.

The accurate reproduction of motion requires that the original 
sampling place the actual information in the correct location. If it 
is between two sample sites, then the Nyquist filtering will spread 
the stimulus over the two adjacent samples.

I have spent years working with high resolution imagery on a variety 
of displays. Bluring is not a desirable thing, as it removes detail 
from the images. But it is critical at lower frame rates to prevent 
motion discontinuity. You would be amazed at how sharp the edges of 
moving object can be on displays with fixed pixel locations when fed 
with high quality source shot at higher FRAME rates. Please believe 
me - the motion portrayal is significantly enhanced, not aliased.

And once again, if you can see individual pixels, you are too close 
to the display.

>
>I'm not saying for sure this is a problem but I have been wondering
>about it recently.  Previously of course I've just been touting the idea
>of fixing the (possible) problem using 1080p displays but they still
>seem slow to arrive.  There are yet no xHD3's and Intel has abandoned
>LCos.  :-(

You are fixated on the wrong issue. The problem is not the display. 
The problem is the sampling. In order to improve the perception of 
detail in the presence of rapid motion you need to increase the FRAME 
rate. At lower frame rates we must blur the moving edges to prevent 
the perception of motion discontinuity.

Regards
Craig
 
 
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at 
FreeLists.org 

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: