"Agamemnon" wrote in message
...

"Roderick Stewart" wrote in message
om...
In article , Ben
wrote:
No one (I think) really addressed the 3rd question regarding colour
resolution. Perhaps, like me, people did not really understand the
question.

I think the OP was referring to the fact that in PAL the vertical
chrominance resolution is only half that of the luminance resolution.
Digital doesn't do that sort of thing.

People who think this is important don't understand the question. All
the analogue colour encoding systems have less than half the horizontal
chrominance resolution anyway, because that's all that's required, so
reducing the vertical resolution to match will do no harm at all. All
these systems are designed to take account of the way the human visual
system works and not waste bandwidth by transmitting information that
nobody will see.

The human visual system does not work like that at all either at close range
or in bright light. The centre of your field of view is more densely packed
with colour receptors than B/W and because of this it provides the highest
resolution image and that's what gives you the ability to read.

I think there are millions more rods and they're more senstive, of the smaller
number of cones only around 2% of cones have blue sensitivity for example.

You can make a lot more valuable use of bandwidth in terms of resolution than
providing a 4:4:4 colour space. I would say compression artefacts or even
curious P frames are a lot more obvious than quantization at 4:2:0.



Az.

"Aztech" wrote in message
...

"Agamemnon" wrote in message
...

"Roderick Stewart" wrote in message
om...
In article , Ben
wrote:
No one (I think) really addressed the 3rd question regarding colour
resolution. Perhaps, like me, people did not really understand the
question.

I think the OP was referring to the fact that in PAL the vertical
chrominance resolution is only half that of the luminance resolution.
Digital doesn't do that sort of thing.

People who think this is important don't understand the question. All
the analogue colour encoding systems have less than half the horizontal
chrominance resolution anyway, because that's all that's required, so
reducing the vertical resolution to match will do no harm at all. All
these systems are designed to take account of the way the human visual
system works and not waste bandwidth by transmitting information that
nobody will see.

The human visual system does not work like that at all either at close
range or in bright light. The centre of your field of view is more
densely packed with colour receptors than B/W and because of this it
provides the highest resolution image and that's what gives you the
ability to read.

I think there are millions more rods and they're more senstive, of the
smaller number of cones only around 2% of cones have blue sensitivity for
example.

You can make a lot more valuable use of bandwidth in terms of resolution
than providing a 4:4:4 colour space. I would say compression artefacts or
even curious P frames are a lot more obvious than quantization at 4:2:0.

OK I've done a simulation taking a scanned image and converting it to
separate H, S, L frames and then applying 50% compression to the horizontal
hue and the stretching it back again and then recombining the images and the
colour artefacts are only obvious if I magnify the image by 2.

If I also did another experiments and compressed the both the saturation and
the hue by the same amount the recombined image shows noticeable shimmering
at the edges of objects but its not as bad as the stuff I have watched on
DVD which on my PC which is completely blurred compared to scans and digital
camera photos at the same resolution. Even a full screen facial portrait
looks blurred. Is something else other than the colour being compressed and
if so by how much ? Is dynamic range compression being applied to the colour
or luminance or saturation. How may bits per pixel are being used.




Az.

"Agamemnon" wrote in message news:419a4fa8$0

I think there are millions more rods and they're more senstive, of the
smaller number of cones only around 2% of cones have blue sensitivity for
example.

You can make a lot more valuable use of bandwidth in terms of resolution than
providing a 4:4:4 colour space. I would say compression artefacts or even
curious P frames are a lot more obvious than quantization at 4:2:0.

OK I've done a simulation taking a scanned image and converting it to separate
H, S, L frames and then applying 50% compression to the horizontal hue and the
stretching it back again and then recombining the images and the colour
artefacts are only obvious if I magnify the image by 2.

So much for filtering.

If I also did another experiments and compressed the both the saturation and
the hue by the same amount the recombined image shows noticeable shimmering at
the edges of objects but its not as bad as the stuff I have watched on DVD
which on my PC which is completely blurred compared to scans and digital
camera photos at the same resolution. Even a full screen facial portrait looks
blurred. Is something else other than the colour being compressed and if so by
how much ? Is dynamic range compression being applied to the colour or
luminance or saturation. How may bits per pixel are being used.

24-bit colour space.

http://tinyurl.com/4wnhq


Az.

"Aztech" wrote in message
...
"Agamemnon" wrote in message news:419a4fa8$0

I think there are millions more rods and they're more senstive, of the
smaller number of cones only around 2% of cones have blue sensitivity
for example.

You can make a lot more valuable use of bandwidth in terms of resolution
than providing a 4:4:4 colour space. I would say compression artefacts
or even curious P frames are a lot more obvious than quantization at
4:2:0.

OK I've done a simulation taking a scanned image and converting it to
separate H, S, L frames and then applying 50% compression to the
horizontal hue and the stretching it back again and then recombining the
images and the colour artefacts are only obvious if I magnify the image
by 2.

So much for filtering.

It depends how you do the compression. I did a pixel resize. With a Bilinear
resample the artefacts are massive and obvious a mile away and this was when
just the hue was compressed. The filters are probably what are causing the
blurring then.


If I also did another experiments and compressed the both the saturation
and the hue by the same amount the recombined image shows noticeable
shimmering at the edges of objects but its not as bad as the stuff I have
watched on DVD which on my PC which is completely blurred compared to
scans and digital camera photos at the same resolution. Even a full
screen facial portrait looks blurred. Is something else other than the
colour being compressed and if so by how much ? Is dynamic range
compression being applied to the colour or luminance or saturation. How
may bits per pixel are being used.

24-bit colour space.

http://tinyurl.com/4wnhq


Az.

In article , Agamemnon
wrote:
What the eye perceives is down to the concentration of receptor cells.

Not entirely. The eye isn't a camera, or a measuring instrument. It has a brain
attached. The combination of the two has evolved to process information that is
significant to their owner, who then judges it subjectively.

systems, i.e all the main broadcast systems, was based on research that
showed
that visual acuity is less for changes in colour than for changes in
brightness. All these TV systems use this presumption and only transmit
chroma

In bright light the eye is more sensitive to colour information and in dim
light it can't see colour at all and relies on peripheral black and white
vision. That's why you are more likely to see dim stars by not looking at
them directly but to the side.

Quite true, but this isn't the point I was trying to make.

On top of that if two colours are placed next
to each other the brain will merge them into another colour if you are
looking from far away.

Which is another way of saying that visual acuity is less for colour changes
than for brightness changes, which *is* the point I was trying to make.

Rod.

The message
from "Agamemnon" contains these words:


"Roderick Stewart" wrote in message
om...
In article , Ben
wrote:
No one (I think) really addressed the 3rd question regarding colour
resolution. Perhaps, like me, people did not really understand the
question.

I think the OP was referring to the fact that in PAL the vertical
chrominance resolution is only half that of the luminance resolution.
Digital doesn't do that sort of thing.

People who think this is important don't understand the question. All
the analogue colour encoding systems have less than half the horizontal
chrominance resolution anyway, because that's all that's required, so
reducing the vertical resolution to match will do no harm at all. All
these systems are designed to take account of the way the human visual
system works and not waste bandwidth by transmitting information that
nobody will see.

Yes, quite true most of the time. That's not to say a matching colour
resolution wouldn't improve the percieved quality on static or nearly
static images. Even so, the colour component bandwidth costs in analogue
transmission systems just don't justify the relatively small improvement
this would make, even considering static scenes.

The human visual system does not work like that at all either at close
range
or in bright light. The centre of your field of view is more densely packed
with colour receptors than B/W and because of this it provides the highest
resolution image and that's what gives you the ability to read. If what you
are saying was done on a computer display it would be completely
intolerable. Its bad enough trying to read coloured text on a dark
background as it is. If a system were not designed with 1:1 colour to pixel
mapping I would not accept it to be Hi-Dfeinition. In fact I would
prefer as
625 line with 1:1 mapping to any 1050 or 720 line system without it. If you
compare a 768x576 DVD screen capture with a scanned image from a
magazine or
digital camera the difference in quality is astonishing.

The most legible type is black against a *white* background, ie maximum
contrast. The colour resolution of the human eye is lower than that due
to luminence differences alone under any lighting conditions.

In moving images, the eye can't percieve the same detail as in a static
image which is why broadcast TV looks so acceptable most of the time
despite it's abysmal standard of resolution compared to even a small
computer display (say 800 by 600 pixels).

Digital transmission methods can potentially be used, in conjunction
with computer grade monitors, to give fast action/low resolution, with
high detail on static/slow action scenes which would be a better match
to the way human visual perception actually works.

The bandwidth requirements would remain constant in that rapid changes
require less data per frame but higher frame rate whilst slow/static
scenes would use higher data( which would include full colour resolution
on very slow/static scenes) per frame at slower frame rate.

HTH

--
Regards, John.

To reply directly, please remove "buttplug" .Mail via the
"Reply Direct" button and Spam-bots will be rejected.

The message
from "Agamemnon" contains these words:


"Roderick Stewart" wrote in message
om...
In article , Agamemnon
wrote:
People who think this is important don't understand the question. All
the analogue colour encoding systems have less than half the horizontal
chrominance resolution anyway, because that's all that's required, so
reducing the vertical resolution to match will do no harm at all. All
these systems are designed to take account of the way the human visual
system works and not waste bandwidth by transmitting information that
nobody will see.

All perfectly true except for those people who compare a TV set to a
computer based display system designed to show high resolution images of
static images, usually text as well as the ability to display moving
images and forget that the primary function of a TV set is to show
*moving* images rather than a high resolution image of fine print.


The human visual system does not work like that at all either at close
range
or in bright light. The centre of your field of view is more densely
packed
with colour receptors than B/W and because of this it provides the
highest
resolution image and that's what gives you the ability to read.

Concentration of receptor cells is irrelevant. I was talking about what
the eye
perceives. Choice of chrominance bandwidth for NTSC, PAL and SECAM colour

What the eye perceives is down to the concentration of receptor cells.

But it's still contrast rather than colour differences that provides
the most detail.

systems, i.e all the main broadcast systems, was based on research that
showed
that visual acuity is less for changes in colour than for changes in
brightness. All these TV systems use this presumption and only transmit
chroma

In bright light the eye is more sensitive to colour information and in dim
light it can't see colour at all and relies on peripheral black and white
vision. That's why you are more likely to see dim stars by not looking at
them directly but to the side. On top of that if two colours are
placed next
to each other the brain will merge them into another colour if you are
looking from far away.

with about one third of the bandwidth used for luminance and they all
work. At
the intended viewing distance you can't see that the colour iunformation
is
less sharp.

Well, that does tend to break down a little under certain conditions
such as when zooming out from a snooker table with red balls on a green
cloth slowly turning brown.

In this case, we have a largely static scene filled with objects that
are familiar to the viewer. The viewer's visual cortex then has the
luxury of time to process the (lack of) detail and for the owner of said
visual cortex to spot the change of colour with respect to previous
experiences of directly viewed snooker tables and red balls on green
cloths.

The fact that they would have also observed a similar change of colour
when directly viewing such scenes of snooker tables at longer distances
tends to be forgotten. To be fair, this 'forgetfulness' is more to do
with the importance of remembering 'truths' rather than spurious effects
due to the limitations of our visual sense.

For the most part. the reduced colour resolution nicely matches our
visual perception and would not be a distraction.

If you wish to invoke intended viewing distance then going by what you have
just said at this distance you cant see the difference between normal
definition TV and high definition TV either so why bother with high
defintion TV, since if as you claim you can only perceive 293 pixels of
colour on a 768 column display then doubling that to 768 pixels on a 1572
column display will have no noticeable effect on effect the colour or even
the intensity definition either.

The eye uses contrast to pick out detail, chroma differences at
identical luminence levels are percieved at lower resolution. The
analogue transmission systems rely on this limitation of visual
perception to save on bandwidth with regard to colour information.

Obviously the fact that people are developing high definition TV shows that
the difference is noticeable which more or less proves that the so-called
research into colour perception you are referring to is a load of baloney
even at the intended viewing distance.

HDTV is a more visually satisfying experience, especially when dealing
with largely static scenes but for scenes with rapid changes it becomes
largely wasted on the eye/visual cortex system that is our sense of
sight. Analogue transmission of our current 625 PAL standard is wasteful
enough of bandwidth just as it is since most of the time it is involved
in repeatedly transmitting the same information over and over at 25
frames a second (50 fields a second).

HDTV is now only becoming feasable due to the emergence of more
intelligent digital transmission techniques which can remove most of the
redundency inherent in analogue transmission systems.

As far as I am concerned the intended viewing distance is me sitting right
in front of my monitor and I can read colour text at 1600x1200 resolution
well enough and I can see the individual colours and I can see the blurring
caused by halving the colour resolution perfectly well and I can't
stand it.
I'd prefer a lower screen resolution without any compromise to the colour
bandwidth than a higher screen resolution with the colour bandwidth halved.

That applies to all of us but you seem to forget that you are dealing
with a *static* scene. If you could run a program that jiggled that text
about, you'd need to increase the point size in order to be able to read
it again. How much you'd need to increase the point size depends on how
much 'jiggle' is being applied. Obviously, the more the jiggle. the
greater the increase in point size to restore legibility.

Don't forget, the sensation of sight is a combination of the eye and
the visual cortex which processes the stimuli. The processing function
has a bandwidth limit which optimises between processing detail in
largely static scenes and processing movement at the sacrifice of detail
when the scene is one with rapid temporal changes.

A digital system can take full advantage of this and go beyond the
simple analogue methods of tailoring the colour content to match the
eye's reduced acuity in this area of perception. A properly designed
HDTV system could give you the resolution you crave for in static scenes
and drop the resolution down to the standard of the current analogue 625
PAL for rapid action content without you noticing the loss.

Since TV broadcasting is largely to do with *moving* images, the lack
of detail only becomes apparent on those relatively few occasions that
involve static scenes which tend to be relatively short lived and
involve close ups when text or hand writing needs to be legible.

The analogue system meets our visual needs very nicely when it comes to
movement but is somewhat lacking in satisfying our need for detail in
static scenes. To satisfy that demand using analogue techniques is
simply far too wasteful of transmission bandwidth and hence,
impractical. Digital methods of transmission will allow both needs to be
met with less consumption of bandwidth.

HTH

--
Regards, John.

To reply directly, please remove "buttplug" .Mail via the
"Reply Direct" button and Spam-bots will be rejected.

"Aztech" wrote in message
...
"Agamemnon" wrote in message news:419a4fa8$0

[snip]

If I also did another experiments and compressed the both the saturation
and the hue by the same amount the recombined image shows noticeable
shimmering at the edges of objects but its not as bad as the stuff I have
watched on DVD which on my PC which is completely blurred compared to
scans and digital camera photos at the same resolution. Even a full
screen facial portrait looks blurred. Is something else other than the
colour being compressed and if so by how much ? Is dynamic range
compression being applied to the colour or luminance or saturation. How
may bits per pixel are being used.

24-bit colour space.

True for MPEG2 4:2:0/4:2:2 distribution (and most other compression schemes
I think), though production kit may well use 30bit (10 bit Y, Cr, Cb is
sometimes used in production environments, and 12 bits plus if film DI stuff
is being done in the HD spatial domain)

Steve

"Agamemnon" wrote in message
...
1) Are Hi-Definition DVD's available and if so can I play them on my DVD
Re-Writer with my present hardware and software ?

There is no consumer standard for High Definition DVDs as such yet.

BluRay (not quite the same as DVD) consumer recorders are available in Japan
to record off-air digital HDTV broadcasts - they can store about 25Gb on a
single layer disc (with around 48Gb on a double layer one I believe)

There is a proposed HD-DVD standard - but I don't think it has launched yet.

Microsoft have released some Windows Media 9 HD material on DVD-ROM - but
this is not a formal standard. In the case of Terminator 2 (the most
mainstream release) it is heavily DRMed - and requires internet connection
to be verified as coming from a US IP address (most well known proxy servers
won't work apparently)

AIUI it is possible that a number of codecs may be supported. MPEG2 (as
currently used OTA - and thus also used by the BluRay recorders on sale in
Japan - which like the HDTV DVHS recorders sold in the US - record the
off-air data without decompressing and recompressing), Windows Media 9, and
MPEG4 (H264 ?) are all likely candidates (and more than one may be
supported)

The DVB (broadcast rather than DVD but obviously these things are linked
these days) organisation has just added H264 AVC to MPEG2 as a DVB video
codec I believe.


2) Does the system use Anamophic Windscreen or true Windscreen ?

Most HDTV schemes proposed use square pixels, 1920x1080 and 1280x720 are the
two 16:9 HD sampling schemes used for production. Many MPEG2 HD broadcasts
currently on-air in the US, Japan, and HD1 in Europe, broadcast stuff with
these sampling structures as well. Australia has chosen to use anamorphic
1440x1080 for transmission (though production is still 1920x1080) to reduce
bandwith requirements (it is still pretty much twice the horizontal and
vertical resolution of 16:9 anamorphic standard definition at 720x480/476)
AIUI some US satellite re-broadcasters are also reducing the horizontal
resolution - though not sure if they do this by filtering (and still
broadcast at 1920 - just with less HF information present), or by subsamping
(and broadcasting at 1440 or lower)

I would expect any imminent HD DVD format to support 1920x1080 and 1280x720
as the main sampling structures. They may also support other schemes.


3) Is the colour of the same resolution as the luminance or is it half the
resolution or worse like PAL ?

I suspect that 4:2:0 sampling will continue to be used - as it has equal
horizontal and vertical chroma resolution (4:2:2 has higher vertical
resolution than horizontal - though this has advantages during production)

I very much doubt a full-bandwith RGB or YCrCb system will be used for a
domestic format - for standard video, at normal viewing distances, the
eye/brain really doesn't see as much chroma information as luminance. There
is loads of research backing up the choice of YCrCb/YUV/YIQ transmission
schemes used - which give more bandwith to luminance than chrominance - this
dates back to the launch of colour TV transmission in the 50s (with NTSC),
and continues through PAL, MAC and MPEG2.

4:2:0 is better than PAL analogue, and much better than NTSC analogue, in
chroma bandwith terms - so there is still a significant chroma improvement.
It really isn't possible to justify moving from a 12 bit system (1xCr and
1xCb sample for every 4xY samples) to a 24 bit system (1xCr and 1xCb system
for every 1xY sample)- doubling the bandwith of the video signal you need to
compress, when the improvements are so marginal. If you didn't increase the
bandwith you'd be massively increasing the compression, adding huge amounts
of artefacts to the system

I would expect H264 and Windows Media 9 to both continue with 4:2:0 chroma
subsampling. (In fact some HDTV format VTRs - like the non-SR HDCam format-
I believe, use 4:1:1 sampling, as does miniDV in 525/60 territories, as well
as DVCPro in all territories)

Steve