MPEG1 is better than MPEG2 for VHS - (x)vcd

[holistic]

Ok OK i now have your attention. This topic is not intended to draw everyone into another M1 vcd - M2 svcd pi$$ing contest. It is just an observation i have made after doing many test files.

CAPTURE - test file - home VHS video - 20 seconds.

I have tried (NTSC) both 352*240 and 352*480 captures.
Yes yes yes there is a slight difference in the 352*480 when resized back to 240 but 'my' eyes tell me it is not worth the hassle (you decide)

ENCODE - using TMPEGenc - test file 352*240 AVI
* audio is 128 mono (vhs dictates this) 44100 hz

Using TMPEGenc I encoded the original sample at MPEG 1 various bitrates up to MAX 2500. Then I encoded original in MPEG 2 at 2500br . [ before you ask : why not higher BR. i want to get NO less than 30 minutes on a CD - 40 preferred ]
Of course the MPEG 1 at 2500 is the better of the MPEG1 test files but when compared side by side with MPEG 2 sample i could not pick a winner.


A NEW TEST
In addition : i also did a 20 second 480*480 AVI capture and encoded in MPEG2 at 2500, 5000, 9000 Br.
The 9000 bitrate MPEG2 was great but it doesn't forefill my 30 + minutes per CD requirements.
When compared - MPEG1 at 2500 (200% size) : MPEG2 (after conversion to (x)SVCD) at 100% . I have decided that the slight "quality" (if you choose to call it that - since VHS tape doesn't have resolutions this high) is not worth all the extra work to get the end results.


CONCLUSION
capture at 352*240 (480 if you feel it makes a difference)
encode at 352*240 MPEG1 2pass VBR, high quality
leave in mpeg1 or make xVCD
I have no use for the multi channel , subtitles etc

.............. comments ......... thoughts ! - opionions

[jerome]

To watch on a computer I agree with you if the source is interlaced. However, for watching the movie on a PAL TV (I'm European and don't know about the result if the source is NTSC), I strongly recommend to capture and encode in higher resolutions. MPEG2 also gives slightly better compression results (although this is also a matter of taste), but what is more important is that it supports interlaced video. For your own camcorder recordings this is very nice.
Therefore I capture at 1/2 dvd PAL format (352x576) and encode in mpeg2 interlaced 2500 kbps. very nice result!

[holistic]

[quote]
On 2001-09-17 17:59:28, jerome wrote:

I strongly recommend to capture and encode in higher resolutions.

-> why ? VHS is only approx (have seen few different spec) 320 * 240 (interlaced) what is achieved by capturing at higher res ? where does the extra video data come from ??

[Hollywood004738]

I capture .avi at 640*480 and use noise reduction. I have found that when encoding the file with De-interlacing on the resulting 352*240 file is really good quality

[holistic]

TRUE OR FALSE


Standardized in 1992, MPEG1 was intended for VHS-quality signal transmission primarily for the then-nascent digital video market and is still considered an efficient use of bandwidth and storage space. MPEG2 was created as the standard for digital broadcasting to provide higher levels of bandwidth transmission needed by, amongst others, direct satellite service (DSS) providers. MPEG1 has an average compression rate of about 1.5 megabits per second (Mbps); the largest possible compression rate for MPEG1 is slightly more than 5 Mbps. MPEG2 bitrates fall between about 3Mbps and about 15Mbps. Interestingly, at bitrates below 3 Mbps, MPEG1 actually performs better than MPEG2. This is because the higher level of precision built into the MPEG2 algorithm requires more processing than MPEG1. At lower bitrates, the percentage difference is great enough to cause significant digital artifacts to appear in MPEG2 encoding that do not appear in MPEG1. MPEG2 should not be used at bitrates lower than 3 Mbps.


copied from : http://www.d-co.com/digicaster-faq.html

Seems to fit in with my theory .

comments , opinions , thoughts ... lets get this sorted out once and for all




<font size=-1>[ This Message was edited by: holistic on 2001-09-18 06:43:54 ]</font>

[SatStorm]

No, I don't agree....

You see, mpeg 2 encoding is not "automatic"
There is no good "load template and hit encode" solution for mpeg 2. No...
With ready templates, you don't get the most of it....
You must set some things manual...

One example:
For mpeg 2 encoding, you always have to de-interlace the video source, with the adaptive de-interlace filter...
That has nothing to do with the nature of the output file, which maybe will be interace too.

I use (as all of as) believe that "Interlace in, Interlace out", for mpeg 2 encoding. Wrong...

Interlace in >> De-interlace with adaptive interlace filter >> Interlace (or progressive scan) out

With this method, you get much less blocks in mpeg 2...

Also, choose "detect scene change on" on TMPEG. That will add full frames to compression. The file gets bigger, but the quality gets MUCH better...
Don't use "no motion search for half pixel". This creates blocks! By doing re-interlace and have detect scene change to on, the blurness eliminates about 80% Not noticable on a TV screen

The "detect scene" feature of TMPEG is bad when you feed the encoder interlace source, you don't re-interlace it and output is again as interlace or progressive...

When I 'll finish my tests, I'll publish 2 templates optimized for re-encoding from satellite feeds and 2 for encoding from AVI source.

[holistic]

A test from VHS tape source is what i would be intrested in most.

tx

[supercrew]

I agree with holistic,

I encode all my VHS captures to mpeg1 and
create xvcd's.

VirtualDub Process:
1) Capture 352 X 480 using Picvideo quality 18,19.
2) Crop 12 lines from bottom to remove jitter.
3) SmartDeinterlace blend
4) Resize to 352 X 240

CCE Process
1) Video Bitrate 1800kbs
2) Audio Bitrate 128kbs
3) Header after every gop

Cut with IfilmEdit 1.4
Burn with Nero

Shows kinda soft, but with no artifacts
on my Pioneer 525.

Get about 40 minutes on cdr.

[kingoflubrication]

satstorm, when you say "de-interlace" the source do you mean within tmpgenc: under the video source settings choose "non-interlace", or is this something completely different? does "de-interlacing" the source require another program like virtual dub, and where can i get the adaptive filter? i would appreciate any help, as i am trying to minimize mpeg2 blockiness.

[acerglenn]

You all have your opinions about SVCD and its worth, but please indulge me while i ask a dumb question. Hope someone can help me - you all seem to be really smart on the topic, so hopefully.

I may buy a card whose maximum capture resolution is 640 x 480 (as it's the best quality, cheapo that have AV inputs).

But, I want to make SVCDs which require a resolution of 480 x 576.

Will the eventual svcd work out all right if i capture at 640 x 480 and resize?

thanks for you help! (ps can you tell i'm a newbie and have no idea whatsoever?)

[SatStorm]

@kingoflubrication
Just load your avi/mpeg/d2v/whatever to tmpegenc, load your favorite template and go to advance menu. Hit "Deinterlace", you gonna see many options. Roll the video to a frame with interlace problems, and use them.
I personaly use Odd/Even Field (field adaptive) option, but some times others like double give me better results...
It is like using Virtualdub filters, but I think de-interlace is better with tmpegenc... The encoding time is almost the same using this filters

@acerglenn
Depends of the source. For VHS, I think is OK (not great, but ok). For SVHS, MPEG or Live TV transmissions, get a card capable for 702X576 grabbing. It gives you much more...

[ktnwin]

I believed I understood the reason why XVCD (352x240, bit rate 2500) is better than SVCD (480x480, bitrate 2500)

The 480x480 is 2.7 times more resolution than 352x240. Just take (480x480) / (352x240) and you will see the ratio.

Since there are only 2500Kbits to update the screen per second, there are 2.7 times more pixels updated for XVCD than for SVCD, therefore there will be less blockiness if any.

However, in term of picture crispness, 480x480 is better than 352x240.

MPEG encoding is simply a game of "updating only the difference from one frame to the next", the higher the bitrate, the more delta portion of the screen can be updated to come closer to the key frame (original picture).

I also note that VHS capture contains lot of noises which show up as delta between frames. Encoder wastes a big part of the bitrate just to update noise.

Just my two cents.

[jerome]

to holistic. remember me? was the first one to reply. what a discussion, jee. well, to explain my opinion I give you the following thoughts.
First of all, I have tried al kinds of de-interlacing tools. All not satisfactory, at least to me.
De-interlace blend on a 352x576 capture gives me effectively a resolution of 352x288 (due to the blending, hehe). Then I'd rather use 352x288 in the first place, this gives no out of focus movie in fast moving scenes.
Smart de-interlacing programs such as the rather good one from Donald Craft still suffer from artifacts, not acceptable to me.
Second, if you don't capture at 480 (NTSC) or 576 (PAL), but at half the resolution you don't get the 50 half frames per second but only the 25 A or B half frames (fields) so you miss 50% of the information. Ok, I know that the VHS resolution is only 200 lines. However, the outputted format of your camcorder or VCR is still 480 or 576 lines which contain 100% of the information (and not 2 times 100% as suggested by you. The A and B fields are NOT the same! They complement each other!).

Enough theoretical fuzz. I did also some tests (by the way I use TMPGEnc):
1. capture at 352x288 and encode 2500 kbps MPEG1 at best settings.
2. capture at 352x576 and encode 2500 kbps MPEG2 at best settings (source and destination interlaced).
3. capture at 352x576 and encode 2500 kbps MPEG1 at best settings (progressive source, e.g. VHS movie 24 frames/sec)
4. the same as 3. but now with MPEG2.

Results:
With interlaced source 1. looks best on a computer screen because you don't have to de-interlace (as said before, I hate the result of current de-interlacing methods);
With interlaced source 2. looks much better on a TV screen(somewhat richer color and noticably sharper);
With non-interlaced source, both 3. and 4. look nice (as I said it is also a matter of taste) and again sharper than 1.

I also noticed that at the same bitrate there is no difference at all between 288 and 576 in terms of compression quality for both MPEG1 and MPEG2 (e.g. blocking does not increase). I assume that this is because MPEG internally is 50 frames per second in both cases. At 288 it just makes a copy of one field to obtain a whole frame, while at 576 you really have both A and B fields of 288 lines each.

Best Regards,
Jerome.

[bky]

greetings to all.
i'm looking for some helps on how to use TMPGEnc/VirtualDub. i dont really know how to use these two. I got assus gForce2 32M V7100 Deluxe Combo as my capture card, running on win98. well my problem is this, i captured using assus Live and also assus virtual VCR from VHS to avi. and use TMPGEnc to encode to mpg. I also had Kystal Codec and Dixv(or is it DivX) codec along w/ assus codec and microsoft codec. i'm not sure what to use cus i dont really understand all these.
i've tried TMPGEnc before it gives me an 'ok' result but i wanted better. NOW its worse, um, the sound is like overlapped or something, that there is like this horrible ehco trailing the sound.
after reading all these, i see that you all know TMPGEnc very well. so i'm hopeing that you can help me out. btw, i use Nero 5.5 to burn
thanks in advance.

[SatStorm]

I've done more tests...

Source: VHS, S-Video cable
Card: Asus 7700 GF2 deluxe

Capture: 720 X 576, uncompressed avi, with asus live program (Virtual dub also works, I prefer default program for the test)

I feed tmpeg 12h with the avi (direct, no frameservers etc)
On the advance menu, I select:
Ghost (0 position, 5 for blurness, auto for brightness)
Sharpness: About 60 - 75 both vertical/orizontal lines
De - Interlace: Odd field (adaptive)
Resolution of the output file: 352 X 576 (CVD Pal), Bitrate 2350Kb/s, Interlace output

With this combination, my mpeg2 is like the VHS source. No blocks, no flickering. Some blocks only to very fast scenes, you can eliminate the effect if you set bitrate about 2600kb/s

When you capture to 352 X 288 and encode to 352 X 576, you get worse picture but smaller file.

I think, de - interlace with tmpeg is better than do the same with vdub....

[jerome]

Hi satstorm. I do not understand why you de-interlace when you want to produce an interlaced output. But on the other hand, I will try your settings because theory is nice but results count and are decisive.

[d4n13l]

I find MPEG1 to be better given lower bitrates (upto 2200Kbits). I normally rip DVDs and encode so that it'll fit on to 2 800MB CDRs and watch on DVD player and 28" W/screen TV. I could never get any MPEG2 file to look as good as MPEG1 upto 2200Kbits (Max rate for many DVD recorders), no matter what settings or methods I tried. I found the higher res MPEG2 to be far too blocky. But thats just what works best for me. My honest opinion is that MPEG1 deals with lower bitrates more effectively, MPEG2 is really for higher bitrates as it was intended for - "big surprise".

P.S. I only use 352x288 XVCD.

[jerome]

hi satstorm.
i incorporated your changes in my settings. sorry, but colors suffer and de-interlacing artifacts appear. For watching on a PC-monitor, your method isn't too bad, but it lacks quality on a TV-screen. I hoped that tmpgenc would de-interlace better than virtual dub and it does, you're right, but I still like the result much more when encoding interlace >> interlace.