PDA

View Full Version : MPEG-4 Capture


VCD MASTER
30th January 2005, 09:42
Hia everyone, I'm putting this suggestions here 'cause a few days ago I've had the most hard test for any capture procedure. I've received from a friend an older VHS tape containing the Pink Floyd's Live Show called "pulse". For anyone who knows, this maybe is the best live show done in the 90's, a very colorfull performance. Well, the VHS tape i got had a lot or tape noise, resulting of storing near to big speakers or somelike that.

I've tried almost every suggestion of this forum and the results were, in most of cases, very poor. Seems to be that the statical noise on tape adds a lot of motion in the image (like a rainy night) and the motion control in capture becomes almost useless, at the same time that colors are changing over and over again 'cause that's the video itself (colorfull, as I've said before).

Capturing in avi through IUVCR using DIV-X (from v4.0 to v5.02) becamed an autentic nightmare, even when I have a Pentium 3.3 ghz processor, 800 mhz bus on board, 256x video card accelerator and a Matrox Pro capture card.

But, the problem finally was fixed using PVR, powerfull tool from Mainconcept, with MPEG-2 codec, VBR from 1800 to 2200 kbps and frame size 352x240. This last is the key factor (I think so), 'cause if you have a poor quality tape, why spend bigger frame size?. The amount of data in capture was around 13-15 mbps and the quality was amazing (capturing all noise) without no one to pixelation.

I've found that mpeg-2 codec doesn't have any considerable quality change if you're capturing at 352x240 and if you use 2200 kbps as data rate, but almost better, this software (PVR) has an amazing sincronicity control, even when processor is working on something else.

I've had finally despited about IUVCR and WINDVR for these kind of captures, now that those works over the operative system itself, not as PVR that works over JAVA, you gotta probe capturing at 480x480 and Q factor at 1 !!!

Ironically, good captures of poor tapes are those which noise amount is equal to the original without pixelate, specially on shadows and low illuminated sections of the image; and this will become a lot harder to denoise the video but is always better denoising over noisy siganl that denoising over noise and pixelated signal.

Finally, I would like to recomend WINVDR as a good choice for capturing mpeg on bizarre frame sizes (448*336, 416*312,so on...) and of course, bit rate increases at the same time, but it has to be done very accurately because the frame-size vs. bitrate relation is not lineal!!! Maybe the bad word is the contrast manipulation trying to decrease the processor's work over black-shadowed objects and motions, but that's not present problem on PVR.

Ok, tks everyone, keep on capturing!!!

Boulder
30th January 2005, 09:58
The best way - regardless of the source - to capture is still HuffYUV, provided you have the diskspace;) Lossless is always lossless and lossy is, um, lossy.

Wilbert
31st January 2005, 10:20
... or mjpeg for long caps.

But, the problem finally was fixed using PVR, powerfull tool from Mainconcept, with MPEG-2 codec, VBR from 1800 to 2200 kbps and frame size 352x240. This last is the key factor (I think so), 'cause if you have a poor quality tape, why spend bigger frame size?
Because you are throwing away half the info (every second field; i'm talking about vertical resolution here).

Capturing in avi through IUVCR using DIV-X (from v4.0 to v5.02) becamed an autentic nightmare, even when I have a Pentium 3.3 ghz processor, 800 mhz bus on board, 256x video card accelerator and a Matrox Pro capture card.
There can be several reasons why it is a nightmare (CPU too slow, DivX doesn't support interlaced stuff). In addition, like i replied a while ago

"No. In general, people use huffyuv/mjpeg/dv for capping, and not without any reason. If, for some reason, you do want to use an "end" mpeg4 codec, you should use ffvfw and not DivX/XviD. The latter doesn't cope to well with noise and motion estimation (link can be found in the analogue capping guide)."

Btw, did you try AviSynth already to clean up your caps?

VCD MASTER
31st January 2005, 19:21
Ok, Boulder, As I can see you're fighting with the same phantom as I was a few months ago.

I you know the statistical behaviour of any codec on capture, there's always an asintotic growing between 90% and 100% q factor, it seems to be that you don't know the mainconcept's development of mpeg-2 codec. If you change Q from 90% to 91% you get a bigger ( and I say bigger ) file size and the difference between them is less than minimal, and even more between 91% and 92%, 92% and 93%,..., and so on.

Wilbert, I'm capturing using Interlaced fields, if you've read my past post quietly I've wrote that my processor is a Pentium IV at 3,3 ghz clock speed, Ok?

See ya, boyz.

jggimi
31st January 2005, 19:30
I've moved VCD MASTER's recent post, and followups, into its own thread, since it doesn't have to do with our guide, but rather, with VCD MASTER's preferred processes.

Boulder
31st January 2005, 20:07
Originally posted by VCD MASTER
Ok, Boulder, As I can see you're fighting with the same phantom as I was a few months ago.

I you know the statistical behaviour of any codec on capture, there's always an asintotic growing between 90% and 100% q factor, it seems to be that you don't know the mainconcept's development of mpeg-2 codec. If you change Q from 90% to 91% you get a bigger ( and I say bigger ) file size and the difference between them is less than minimal, and even more between 91% and 92%, 92% and 93%,..., and so on.

Well, you don't seem to understand what lossless means:) MPEG-2 is very much lossy which means that you do not get what you see. With lossless codecs what you see is what you get. With lossy codecs the next generation is always worse than the one where it came from.

Don't get me wrong, I capture in MPEG-2 but I would still use HuffYUV if I could. The board just doesn't do AVIs.


frame size 352x240

VHS has two fields so the appropriate resolution would be 352x480. That's what Wilbert was after.