352 vs 720 resolution at 2700 Kb/s

[BrainStorm69]

OK, even though it's slightly off topic - here are 3 images of a test pattern capped off of VHS tape. In one case, I captured to avi at 704x480, encoded to mpeg 352x480, and then lanczos3 resized to 704x480. In one case, I captured to avi at 352x480, encoded to mpeg 352x480, and then lanczos3 resized to 704x480. In one case, I captured to avi at 704x480, encoded to mpeg 704x480.

The files are in no particular order. Which is which? (btw, each is about 1MB in size)

****EDIT****

I'm changing the question. Please rank them from highest to lowest in perceived resolution.

sample01.tga

sample02.tga

sample03.tga

[FulciLives]

I will take a stab at it:

Pic 1 = I captured to avi at 704x480, encoded to mpeg 352x480, and then lanczos3 resized to 704x480

Pic 2 = I captured to avi at 352x480, encoded to mpeg 352x480, and then lanczos3 resized to 704x480

Pic 3 = I captured to avi at 704x480, encoded to mpeg 704x480

This is based on my eyes saying thta Pic 3 looks the best and Pic 2 looks the worst.

- John "FulciLives" Coleman

[johns0]

Sample 1=pic3
Sample 2=pic1
Sample 3=pic2

[BrainStorm69]

@ Fulci - since you based this on which "looks" better, then let me ask the question somewhat more precisely and see if your answer changes - how do you rank them in terms of most resolution to least?

[FulciLives]

@ Fulci - since you based this on which "looks" better, then let me ask the question somewhat more precisely and see if your answer changes - how do you rank them in terms of most resolution to least?

Most to least resolution:

1.) PIC 3
2.) PIC 1
3.) PIC 2

PIC 3 has the most and PIC 2 has the least.

BTW leaving now for some holiday shopping even though it's nearly 9pm EST.

Go figure.

- John "FulciLives" Coleman

[BrainStorm69]

Best time to shop

[mpack]

Mpack - reference to #1, analog doesn't really have a resolution, but we can speak of equivalents.

Someone said that already - and it isn't quite correct. Analog tv has no specific horizontal resolution (ie. lines are not made up of a fixed number of discrete samples), but analog tv definitely does have a fixed vertical resolution: a tv signal is made of a specific number of discrete lines (though not all contain picture), so the vertical sample rate is fixed.

It is generally accepted that VHS "equivalent" resolution is somewhere below 352x480.

That's one of the things I'd like clarification on from a VHS expert -"generally accepted" doesn't mean much to me I'm afraid, if those doing the general accepting are basing their ideas on reading forums such as this one: going back to fundamentals I can visualise a tv signal directly driving the magnetic field of the write head: what limits the effective resolution if there are no intermediate A/D and D/A stages?

The factor that overrides all this concern about source res lies in the nature of how most capture hardware works. They use a FIXED resolution. That's it, done deal.

Hmm. They might use a fixed sample rate, but not a fixed resolution. I've helped design analog video capture hardware, though this was a few years ago now, and not for a PC (I should say that I'm a softie, not an electronics engineer, but as the software engineer who was going to be writing the drivers I was in charge of and intimately involved in specifying how the capture pipeline needed to work). The front end involved a high speed A/D converter driven by a crystal running at (from memory), 27MHz. That clock is high enough for any capture resolution we care about here, and you could scale it down for slower sample rates. Below the standard rate you would of course filter to reduce aliasing noise. The point I'm getting to is that I think your notion that capture at anything other than the "standard resolution" needs to involve fancy scaling hardware is probably incorrect: I don't know the details of ATI's capture electronics, but I'd be rather surprised if it was inferior to the simple design I worked on in 1998.

Why do you say these things about PC capture cards? What is your source of information?

[mpack]

The files are in no particular order. Which is which? (btw, each is about 1MB in size)

I can't tell the difference: but then I wouldn't necessarily expect to, after they've been filtered, then mpeg'ed at an unknown quality setting, then decoded and scaled.

[junkmalle]

OK, even though it's slightly off topic - here are 3 images of a test pattern capped off of VHS tape...

A much more interesting test: use that same test pattern but pan it slowly from left to right at about 5 pixels per second -- about 1/6 pixel per frame. Of course, since you're starting with a digital image and digitally panning it you will get artifacts. But in the area of interest (~225 lines) the video will be pretty clean.

Or, if you don't trust the digital panning, print the test pattern out, pan manually, and record it onto VHS tape via an analog video camera.

Finally, capure the VHS tape at 704x480 and 352x480 MPEG. Burn to a DVD and note the difference between the two caps and the VHS tape. You may find it interesting.

[jimmalenko]

I just *love* it when the tech-heads trade blows over theory

FWIW, I concur with the two johns - Sample 3, Sample 1 then Sample 2, on visual appearance alone.

[MrMoody]

That test pattern is designed to read resolution. Picture 3 is the best at about 240 lines, 1 and 2 are virtually identical at around 230, with 2 having a very slight edge but also more noise. Look at the 3 MHz circle on the top left. 3 MHz is the bandwidth of VHS.

Obviously, 3 was captured at 704, encoded at 704.
2 was captured at 352, encoded at 704.
1 was captured at 704. encoded at 352.

And the test patterns show conclusively that 352 pixels is very close to, but slightly less resolution than VHS. Thanks, now I know not to use 352 on anything from VHS or better.

Post one captured at 352, encoded at 352. I think it will have the least noise of all, but less resolution than 3. If so, it becomes a tradeoff of resolution vs noise, and the noise will get worse when bits are used for motion. The answer is to use a higher bit rate of course, DVD-Rs are cheap.

[lordsmurf]

Tests should all be resized to 640x480 (4:3)

[FulciLives]

Tests should all be resized to 640x480 (4:3)

Actually this is one test where is doesn't make sense to resize to 640x480 because we are testing 704x480 vs 352x480 and since these are just resolution charts and since all were resized to the same 704x480 resolution ... well there is just no need for a proper 1:1 computer screen pixel resolution of 640x480

Doing that (resize to 640x480) would just muck things up in this particular situation.

- John "FulciLives" Coleman

[lordsmurf]

I disagree. We watch video 4:3, not 12:11 or whatever it is.

[FulciLives]

I disagree. We watch video 4:3, not 12:11 or whatever it is.

But the aspect ratio has NOTHING to do with comparing the RESOLUTION and since the test is 704x480 vs 352x480 I see NO SENSE in a resize to 640x480

Why is that SO HARD to understand ???

I just can't grasp how aspect ratio is an issue here. Why can't I grasp that concent in this situation? Because it doesn't apply :P

- John "FulciLives" Coleman

[junkmalle]

If you want to see it at 4:3 aspect ratio burn it to a DVD and watch it on TV.

[mmasw]

Trying encoding a football game at 720x480 at 2700kbps they you will see the flies following the football players in fast action scenes.

flies??....oh....you must be a 49er's fan

[FulciLives]

Trying encoding a football game at 720x480 at 2700kbps they you will see the flies following the football players in fast action scenes.

flies??....oh....you must be a 49er's fan

He was talking about a specific form of MPEG compression artifact commenly called MOSQUITOE NOISE. I can see how someone might also refer to it as "flies".

I don't know a thing about sports and could care less but I take it that the 49er's either suck for real or you just don't like them for whatever reason.

I don't even know what city they are in

The only team I know for sure is that the Steelers are in Pittsburgh. That's because I grew up and live here. Otherwise I have no idea what team is what or who the players are etc.

People here in Pittsburgh keep referring to one of our football players as BIG BEN and there are even BIG BEN t-shirts everywhere all of a sudden ... apparently he is a new team player member or something. I don't have a clue what that is all about except that I guess his real first name is Ben. I have no idea what the BIG refers to otherthan perhaps his penis size perhaps?

Sports like football and baseball and hockey etc. are all just silly nonesense. They are meaningless and lack any aesthetic artistic value.

Not sure how that turned into a rant against "organized" sports but it did

- John "FulciLives" Coleman

[vhelp]

I'm a little confused, *and* a little surprised with your test scenario.
The method(s) you used to produce these pics leave me

I wanted to give an honest pic at the pics, (thought they all do look alike)
but the problem here, to me, is that this test scenario is indeed invalid (imo)
because your test is not clear (the techniques you used to resize and the
mpeg encoding is all so confusing; etc etc)

I don't think that it matters at all when you resize *pics* and produce them
here, and in various resize approaches. Its not making any sense to me.
.
The 704 vs. 352 is almost no contest. Only because of the resizing techniques
used - Interpolating lines.. ususally during the capturing phase internally
by the card.
.
And, depending on your capture card, this could be fixed. If your given card
is actually capping at 352 (and internally interploating them to 704) then your
test scenario is indeed invalid. The unhidden truth to this, is the fact that
the pics all look the same. (w/ slight blemish here or there, no doubt due to
the software's resizing algorithem) They all look pretty much the same. And,
almost all Interpolating resizing techniques will produce the same results.
.
But, what has to be found out, *trueth'wise" is the *TRUE* internal workings of
the capture card. What "resolution" is it finallizing to, when it captures
the video and pushes it onto this container (be it 352 or 704 or 720 or whatever)

In the end, it will always boil down to the encoder, and how well it distributes
the bitrates. It boils down to TMPG vs. CCE v2.50 (I've tried ProCoder and MC,
but in my test scenarios, they two failed to come close. For me, it's TMPG
hands-down for the encoder.
.
Over the weekend, I did some test scenarios w/ 352 encodings. I found that even
4000 bitrates (CBR) was not enought for a 352 x 480 encode. I found heavy pixel
ation everwhere's (course, my source is Antenna, so that doesn't help much, but
the source is comparitable to VHS recording in EP mode)
And, what I did, was I up'ed the bitrate in 1k chunks, until I was satisfied w/
the lesser pizelation. In all honesty, I had to up the bitrate to 9000 to get
a clear non-pixelated video
(mind you, those test scenarios were based on an IVTC source for maximum quality)
.
From that test, I was pretty satisfied with my 352 x 480 encodes. They looked
very good IMO, and no VBR (2pass or more) would have done it for me here with
this test scenario (given my source) It took 9000 CBR to get there for this
resolution. Except for a little fuzz at this resolution, it was the next best
alternative to DVD projects. However, I my future goals are DVD/Large Screen
TV's (maybe widescreen someday) and 720 is the only way to go. With a clean
source, that should not be a problem. However, w/ my noisy Antenna, there's
room for dis-believe. But I can work with it and comprimise for the time being.
.
In the 352 vs. 704 bitrate of 2700 argument, (and after my own tests w/ bitrate
on this) I would have to say the given these two resolutions, I would go with
the 352 as my final res. VHS source *is* clean. It's not noisy, as we (I)
once thought. It's limited in detail, which *could be* perceived as noise.
Then, there is the capture card. And, in my *many* dealings with them, I have
had nothing but Line Noise in them. I can't speak for everyone else here, but
I suspect that everyone else has this too, but they *always* incorporate a
noise or bitrate savings' technique filter in there, that they never see this
noise that I've ben commenting on for years.
But, the noise (rather, lack of) that I'm talking about in VHS is not those
that were recorded to, but those commercial movies. Everyone has to a varying
degree, noise in the Cable or Satellite signals. So, there is bound to be
some noise exhibited during the recording phase.
.
I can live w/ some noise in my source. (so long as I don't *filter* it out,
and what.. taint the sources true orginality? ..no thanks)
Plus, this *is* my hobby. So, I have plenty of room for grace. Anyways.

@ brainstorm69

I say again, that your given capture card is yielding great results, as
evident in the many pics you provided to us all here. You obviously have
no problem in the capturing department. Your system setup is well tuned,
from my point of view. I don't know what other issues you might have, w/
respect to your setup all the way to the encoding to dvd authoring.. other
than the "black level" issue (in another thread) that I'm also dealing
with too. But, you have a good setup thus far, and I commend you for it

@ fulci

You too, seem to have a good setup with your given card. I've seen your pics
also. And, w/ respect to the above.. applies to you too. I don't know if
you suffer from a "black level" issue (dvd to vhs recording, then capture from
vhs) but.. long live Ananlog capturing to you too

-vhelp

[lordsmurf]

The test is invalid to me.

Good job BrainStorm, for doing this. It's nice to see these kinds of things brought to light. However, for the picky-picky that people are doing, this test won't work.

When you compare 704x480 to 352x480 stretched out to 704x480 ... it's a no brainer ... doubling pixels to from 352 to 704 will often (but not always) make the 704 look like shit. Duh. The Photoshop (or editor's) bicubic resize destroyed the image. A DVD player does not resize, it uses the rectangular pixels provided. The PC screen needs 1:1, so our resize is not true.

We watch 4:3 ... The 704 is resized down to this frame size, the 352 is sized up to this. The source determines the perceived resolution. A old VHS tape is going to look the same (unfiltered), regardless of 704, 720, 352. It's below all of these, in an analog equivalent.

A test needs to be 4:3, not a long rectangle, not 8:9.

Add to the fact the Photoshop or whatnot is playing with pixels and this is all a waste of time, IF YOU WANT TO NITPICK LIKE WE'RE DOING. For a general show, woo-hoo, great.

Show this to the average guy on the street, they'll all look the same to him, which is pretty much the point to prove: capturing high res from a low res source is a waste of your time and resources.

It'd be worthwhile only as video clips for download, burned to a disc and played on a NORMAL PERSON'S TV SCREEN (not some uber-nerd videophile $$$$$ 60" screen, as those people usually see things that do not exist).

Sorry to be a party pooper, but I strive to be the voice of reason, as many well know.

This thread has sooooo strayed from post #1.

[tedkunich]

That's one of the things I'd like clarification on from a VHS expert -"generally accepted" doesn't mean much to me I'm afraid, if those doing the general accepting are basing their ideas on reading forums such as this one: going back to fundamentals I can visualise a tv signal directly driving the magnetic field of the write head: what limits the effective resolution if there are no intermediate A/D and D/A stages?

<geek speak>

Since we are talking an analog -> magnetic recording medium, there are lots of variables that affect the bandwidth - tape head gap (the larger the gap, the less bandwidth), bandpass filters in the electronics (slope and cuttoff freqencies), and the size of the magnetic particles on the media itself. Limiting bandwidth will reduce the amount of fine detail information that is rasterized on the screen - not "exactly" the same as reducing "resolution", but in practical applications you get the same effect. As to the resolution number that gets thrown around here, that is all quess work and is at best only an accurate estimation from one set of hardware/media combinations - YMMV.

</geek speak>

T

[lordsmurf]

It's been a long time since I heard people talk about bandwidth as it relates to particles on the tape.

[MrMoody]

We watch 4:3 ... The 704 is resized down to this frame size, the 352 is sized up to this.

Resized down? Up? Huh? How does the DVD player know how big my TV is? What does 704 get resized down to? Certainly not 640, that would cut down the resolution of DVDs. All 704 pixels are output. Analog TVs don't have pixels. All DVD players have analog output.

What it really does is convert the digital pixels to an analog signal at a MHz rate which gets 704 (or 480 or 352) of them done in the time allowed for the scan line to go from one side to the other. In NTSC, 720 pixels are converted at about 11.4 million samples/second to match the scan rate of your TV (as defined by NTSC standard). 704 should theoretically be 11.0 million but I believe most players still use the 720 rate and add narrow black bars to the sides (not visible in the overscan). So really 704x480 is displayed at 3.91:3, not 4:3! Assuming your TV was exactly 4:3 in the first place.

Which brings me to my point, Fulci is right, aspect ratio is irrelevant to resolution: Ratio is just that, a ratio of horizontal SIZE (not pixels) to vertical SIZE (not lines). Pixels are only square if you display them that way. In analog, you can easily stretch or shrink them by changing the deflection in your picture tube. If you want to see those samples at 4:3. just crank down the width on your monitor until it's 4:3. After all that's, the same way your DVD player and TV are doing it.

[BSR]

Those 3 test captures(without zooming in) all look pretty much the same to me.
I have done my own tests comparing captures of 720x480 vs 352x480, which was home video footage. I can see the difference and prefered the crispness of 720x480.
But if I was limited to 2700 kb/s, it would be a no brainer for me, 352x480 2pass vbr.

Fulci -
49ers suck this year, only won 2 games, that's what Mmasw was referring to. They are in SanFrancisco.
Big Ben - New rookie Steelers Quarterback. He started playing the 2nd week of the season and he hasn't lost a game yet. BigBen because he's 6'5" 241 pounds, that's a big quarterback.

4:3 -
Not technically sure about this, but I think creating 4:3 images of these for computer viewing would distort the images further and would contribute another resizers distortions into the image.

[BrainStorm69]

Just to clear up the test methodology - captured to avi with VirtualVCR - encoded to mpeg with TMPGEnc at 2700 kbps CBR - resized the whole mpeg file with Virtualdubmod using Lanczos3 method to new avi file - snapped a source frame from the resized avi file to .tga (lossless) format.

The main thing I wanted to test - which is consistent with this thread - is whether the 704 capture would show greater resolution/detail than the 352 capture. I think it does show that.

The other thing I was testing, since Nelson37 was so insistent that capturing at 704 and software encoding to 352 was ALWAYS superior in terms of detail to just capturing at 352 (which we know it is with most cards), was whether it is true with my particular setup (which has the Philips SAA7130). I think the tests show it's not ALWAYS true.

Just for eveyone's information: Cap1 was captured at 704 then encoded to 352 then resized to 704. Cap2 was captured at 352, encoded to 352, and resized to 704. Cap3 was captured at 704 and encoded at 704.

To be honest, I wasn't sure if resizing to 704 would be the "correct" final resolution. Perhaps as the last step I should have resized to 680, or some other resolution. But I don't think it matters that much in the end analysis - if I had done that, I tihnk the results would have been the same, although perhaps a bit harder to determine.

Now, to relate it all back to the original question, I think the tests show that IF you can set the bitrate high enough to avoid macroblocks and still fit your video onto a DVD, then capturing at 704 (or 720) will actually result in a somewhat sharper picture than 352, even for VHS tapes. But if you can't use a high enough bitrate to avoid macroblocks at 704/720, then IMHO use 352 for your final resolution - the trade-off of a little sharpness for no macroblocks is worth it.

BTW, I always enjoy a little healthy discussion of posts

[FulciLives]

Just for eveyone's information: Cap1 was captured at 704 then encoded to 352 then resized to 704. Cap2 was captured at 352, encoded to 352, and resized to 704. Cap3 was captured at 704 and encoded at 704.

This seems consistent with what I said.

I said Cap 3 was the best looking in terms of resolution. This was the one that was captured at 704 and encoded at 704. This is what I would expect i.e., that it would look the best.

I said Cap 1 was the second best looking in terms of resolution. This was again what I expected since it was captured at 704 but then encoded at 352 before being resized back up to 704.

I said Cap 2 looked the worse in terms of resolution. Again that makes sense because this was captured at 352 then encoded at 352 then resized up to 704.

I am impressed I must say that I got it correct with my naked eye(s).

I kick ass !!!

Thank you.

- John "FulciLives" Coleman

[tedkunich]

It's been a long time since I heard people talk about bandwidth as it relates to particles on the tape.

Just trying to be thorough.... cheap tapes==larger particles==less flux density

[mpack]

Limiting bandwidth will reduce the amount of fine detail information that is rasterized on the screen - not "exactly" the same as reducing "resolution", but in practical applications you get the same effect.

Thanks for your answer.

However, I'm still left somewhat puzzled. You see, I know just enough about analog video signals to be dangerous: for example I know that in color video the chrominance information is encoded as high frequencies superimposed onto the luminance signal. I also know that the apparent detail in an image comes mainly from the luminance component (the eye is less sensitive to high spatial frequences in color, which is why chominance can be subsampled to a greater degree in standards such as JPEG and MPEG).

So, it seems to me that if the bandpass filters were to be strong enough to remove significant spatial information from the underlying luminance signal, it would also screw up the colors. Hmm... maybe it does exactly that - is this what causes chroma noise I wonder?

[Garibaldi]

An observance of 720 being "sharper" is DIRECTLY related to the hardware, software and source. You cannot surpass the resolution of the source (maybe if you use a detailer, but that's a whole different conversation). Most sources are at, under or around 352x480. Almost nothing aside from DV is available to the consumer in that higher res.

I agree. All the capturing I do is from either VHS or 8mm, so the quality isn't that great to begin with.... so 352 works for me.

[vhelp]

@ brainstorm69

w/ respect to ...

Now, to relate it all back to the original question, I think the tests show that IF you can set the bitrate high enough to avoid macroblocks and still fit your video onto a DVD, then capturing at 704 (or 720) will actually result in a somewhat sharper picture than 352, even for VHS tapes. But if you can't use a high enough bitrate to avoid macroblocks at 704/720, then IMHO use 352 for your final resolution - the trade-off of a little sharpness for no macroblocks is worth it.

I liked it, and agree (of course) I can't say for sure weather or not the
352 would be any worse for *larger* screen tvs (specially large widescreen)
and perhaps thats another topic for discussion but it would be nice to
know for sure. However, I feel (or theorize) that it won't be worth it to
much in the long long (vs. 720 encodes) But I guess it doesn't matter all
that much anyways (for the time being)

I kick butt too ... so does brainstorm69
-vhelp
ps: I'm watching Wizard of Oz