Tweet
Robert:
Let me take a shot at your questions. I think you are on the right track,
but I detect some areas where you may be confused. (Join the club :-) !!)
Again, I think this material may be of general interest to the group, so
I'll post it to the list. I think we've got a good "thread" going here,
perhaps others will jump in with their ten cents.
You are right about the loss of resolution. This is a trade-off issue
usually addressed by the manufacturer of video-based motion measurement
systems.
Let's make sure you understand the concept of video frames vs video
fields. First the history lesson. When TV was first created in the 30's,
they tried to paint a sufficient number of lines on the TV screen to give
a reasonable image, but by the time they got to the bottom of the image
the top had decayed. (The phosphors were not up to the job at that time.)
The result was a consistent vertical rolling caused by the image fading in
and out. This was unacceptable, so they decided to draw the image in two
passes--first the odd lines, then the even. The two passes became known
as the "odd" and "even" fields--which constitute a single video frame.
Each of the two field was written in 1/60th second, thereby creating a
frame in 1/30th second. Such images are now described as "interlaced,
with a 2:1 interleave". Today, we have come a long way since the original
broadcast standards--witness the high-resolution, non-interlaced monitors
now widely used in the computer industry. But unfortuantely, were still
stuck with the same old broadcast standards. (Hang in there, they are
working on it, but it's been difficult to get the whole world to agree on
something!! Digital High-Definition TV, which looks like 35-mm slides in
motion, is on the way.) Digital cameras have been built for high-speed
photography and machine vision purposes, but they are far from a
"consumer" commodity--so they cost a lot more. But back to our fields and
frames.
Anyone building a motion analysis system designed to use video input which
complies with one of the old broadcast standards can decide to analyze
each FRAME, or each FIELD. If they do the former, they are really looking
at TWO images which were originally separated by 1/60th second. The
arguement for this approach is that it is possible to obtain greater
spatial resolution because the second image was taken between the lines of
the first. However, the second image was taken after the first, and if
things are really moving around, it is justifiable to question the merit
of trying to reassemble something that was never one in the first place!!
Alternatively, a motion analysis system can look at the individual FIELDS.
As you correctly observed, this reduces the VERTICAL resolution of the
image by a factor of two, but it increases the temporal resolution from
1/30 sec to 1/60 sec. Given the choice, most end-users would rather have
the improved temporal resolution in trade for the loss of spatial
resolution. So business being what it is ( ;-) ) most motion measurement
systems are made this way.
Now regarding your (apparent) confusiuon of the term resolution as it is
used in the video world. Given the fact that (in the broadcast world)
the number of vertical lines in an image is fixed, the VERTICAL
resolution is effectively fixed--by the video standard being used (NTSC,
EIA RS-170, PAL, SECAM, etc.) So the manufacturers can't really compete
in this area. But they can compete in terms of the HORIZONTAL resolution
of their equipment. That is, they can claim to be able to produce
greater resolution within each video line. So you will find
specifications for video equipment claiming to have "400 lines of
resolution". You can read this as ... "Our equipment can reproduce 400
black and white vertical stripes (or 800 transitions) from one end of
EACH HORIZONTAL LINE to the other." You'll find such specifications
associated with video recording equipment and cameras--and other, more
esoteric equipment. Typically, a VHS signal can reproduce about 240
lines, S-VHS and Hi-8 can produce about 400 lines. More sophisticated,
"professional" formats can exceed these specifications.
In response to your question ... "what can framegrabbers do?" The answer
is ... "just about anything you want them to do!" You just have to find
the right framegrabber! Like anything else, you have to find the right
product for the task--or build your own. This is not intended to be a
smug remark, it's just the reality of the situation. Some issues can be
addressed by commercial or custom software written for the particular
framegrabber, other issues depend on the architecture of the hardware. In
a word, you have to ask the vendor the right questions. Obviously,
knowing the right questions to ask is a major part of the problem! I hope
this has begun to fill in some of the blanks for you.
Best wishes,
Jim WALTON
Chairman, SPIE Working Group on High-Speed Photography & Photonics
************************************************** *************
************************************************** *************
* * *
* JAMES S. WALTON, Ph.D. * *
* President * INTERNET: Jim@4DVideo.com *
* 4D VIDEO * *
* 3136 Pauline Drive, * *
* SEBASTOPOL, CA 95472 * *
* * COMPUSERVE: 72644,2773 *
* PHONE: 707/829-8883 * *
* FAX : 707/829-3527 * *
* * *
************************************************** *************
************************************************** *************
On Mon, 22 May 1995, Robert S. Cargill II wrote:
> Jim,
>
> I have a question for you. In your recent, very informative post (I had
> never heard of 3-2 Pull Down), you said that in a 1 to 1 transfer of film
> to video, there is one field per frame of film. My question has to do
> with resolution. If you only get a field for each frame, then your
> vertical resolution is half of what you would get if they did a frame to
> frame transfer. With VHS, this means that you get about 240 lines of
> resolution (which I like to call vertical, but the video industry seems
> to call horizontal
. Is it possible for them to scan the single film
> frame with two offset fields? And if it is, can a normal frame grabber
> board get both fields and assemble them into a single digital image? I'm
> sorry if these are strange questions, but I've always wondered what the
> deal was
>
> Thanks in advance,
>
> Bob Cargill
>
> --
> ************************************************** ***********************
> * Robert S. Cargill II, Ph.D. * Use OS/2, for the fun of it! *
> * rcargill@biochem.dental.upenn.edu * *
> * Post-doctoral Fellow * Cellular and Tissue *
> * PGP Public key available by finger * Biomechanics *
> ************************************************** ***********************
>
>
Let me take a shot at your questions. I think you are on the right track,
but I detect some areas where you may be confused. (Join the club :-) !!)
Again, I think this material may be of general interest to the group, so
I'll post it to the list. I think we've got a good "thread" going here,
perhaps others will jump in with their ten cents.
You are right about the loss of resolution. This is a trade-off issue
usually addressed by the manufacturer of video-based motion measurement
systems.
Let's make sure you understand the concept of video frames vs video
fields. First the history lesson. When TV was first created in the 30's,
they tried to paint a sufficient number of lines on the TV screen to give
a reasonable image, but by the time they got to the bottom of the image
the top had decayed. (The phosphors were not up to the job at that time.)
The result was a consistent vertical rolling caused by the image fading in
and out. This was unacceptable, so they decided to draw the image in two
passes--first the odd lines, then the even. The two passes became known
as the "odd" and "even" fields--which constitute a single video frame.
Each of the two field was written in 1/60th second, thereby creating a
frame in 1/30th second. Such images are now described as "interlaced,
with a 2:1 interleave". Today, we have come a long way since the original
broadcast standards--witness the high-resolution, non-interlaced monitors
now widely used in the computer industry. But unfortuantely, were still
stuck with the same old broadcast standards. (Hang in there, they are
working on it, but it's been difficult to get the whole world to agree on
something!! Digital High-Definition TV, which looks like 35-mm slides in
motion, is on the way.) Digital cameras have been built for high-speed
photography and machine vision purposes, but they are far from a
"consumer" commodity--so they cost a lot more. But back to our fields and
frames.
Anyone building a motion analysis system designed to use video input which
complies with one of the old broadcast standards can decide to analyze
each FRAME, or each FIELD. If they do the former, they are really looking
at TWO images which were originally separated by 1/60th second. The
arguement for this approach is that it is possible to obtain greater
spatial resolution because the second image was taken between the lines of
the first. However, the second image was taken after the first, and if
things are really moving around, it is justifiable to question the merit
of trying to reassemble something that was never one in the first place!!
Alternatively, a motion analysis system can look at the individual FIELDS.
As you correctly observed, this reduces the VERTICAL resolution of the
image by a factor of two, but it increases the temporal resolution from
1/30 sec to 1/60 sec. Given the choice, most end-users would rather have
the improved temporal resolution in trade for the loss of spatial
resolution. So business being what it is ( ;-) ) most motion measurement
systems are made this way.
Now regarding your (apparent) confusiuon of the term resolution as it is
used in the video world. Given the fact that (in the broadcast world)
the number of vertical lines in an image is fixed, the VERTICAL
resolution is effectively fixed--by the video standard being used (NTSC,
EIA RS-170, PAL, SECAM, etc.) So the manufacturers can't really compete
in this area. But they can compete in terms of the HORIZONTAL resolution
of their equipment. That is, they can claim to be able to produce
greater resolution within each video line. So you will find
specifications for video equipment claiming to have "400 lines of
resolution". You can read this as ... "Our equipment can reproduce 400
black and white vertical stripes (or 800 transitions) from one end of
EACH HORIZONTAL LINE to the other." You'll find such specifications
associated with video recording equipment and cameras--and other, more
esoteric equipment. Typically, a VHS signal can reproduce about 240
lines, S-VHS and Hi-8 can produce about 400 lines. More sophisticated,
"professional" formats can exceed these specifications.
In response to your question ... "what can framegrabbers do?" The answer
is ... "just about anything you want them to do!" You just have to find
the right framegrabber! Like anything else, you have to find the right
product for the task--or build your own. This is not intended to be a
smug remark, it's just the reality of the situation. Some issues can be
addressed by commercial or custom software written for the particular
framegrabber, other issues depend on the architecture of the hardware. In
a word, you have to ask the vendor the right questions. Obviously,
knowing the right questions to ask is a major part of the problem! I hope
this has begun to fill in some of the blanks for you.
Best wishes,
Jim WALTON
Chairman, SPIE Working Group on High-Speed Photography & Photonics
************************************************** *************
************************************************** *************
* * *
* JAMES S. WALTON, Ph.D. * *
* President * INTERNET: Jim@4DVideo.com *
* 4D VIDEO * *
* 3136 Pauline Drive, * *
* SEBASTOPOL, CA 95472 * *
* * COMPUSERVE: 72644,2773 *
* PHONE: 707/829-8883 * *
* FAX : 707/829-3527 * *
* * *
************************************************** *************
************************************************** *************
On Mon, 22 May 1995, Robert S. Cargill II wrote:
> Jim,
>
> I have a question for you. In your recent, very informative post (I had
> never heard of 3-2 Pull Down), you said that in a 1 to 1 transfer of film
> to video, there is one field per frame of film. My question has to do
> with resolution. If you only get a field for each frame, then your
> vertical resolution is half of what you would get if they did a frame to
> frame transfer. With VHS, this means that you get about 240 lines of
> resolution (which I like to call vertical, but the video industry seems
> to call horizontal
. Is it possible for them to scan the single film> frame with two offset fields? And if it is, can a normal frame grabber
> board get both fields and assemble them into a single digital image? I'm
> sorry if these are strange questions, but I've always wondered what the
> deal was
>
> Thanks in advance,
>
> Bob Cargill
>
> --
> ************************************************** ***********************
> * Robert S. Cargill II, Ph.D. * Use OS/2, for the fun of it! *
> * rcargill@biochem.dental.upenn.edu * *
> * Post-doctoral Fellow * Cellular and Tissue *
> * PGP Public key available by finger * Biomechanics *
> ************************************************** ***********************
>
>