DV to H264 at 3100kb/s and still has macroblocks?

[SatStorm]

OK, I decided to play a bit with H264, which I use a lot these days.

So, I used my Canopus A100 to capture a VHS tape to DV and I convert it to H264.

The source is 720x576 interlaced DV and the result is 720x576 H264 interlace. I used Virtualdub and the latest unofficial Vfm x264 for my tests. Default x264 settings, CBR, 3100Kb/s.

Surprise: I have macroblocks! A few, but they are there...
So, what I'm doing wrong?

The original VHS is 20+ years old, full of any possible kind of noise. I don't wish to use filters just to keep it that way. The point is to make a 1:1 digital copy of my VHS tape to H264.

I have great results using Full D1 and 1150kb/s but - of course-, those results are far from "perfect". There are excellent and with filtered source, DVD-Like in quality. So, I thought 3100 would be enough for a noisy VHS. Obviously it is not.
In your opinion (and on your tests), what is the macroblock-free bitrate for H264 at Full D1?

Thanks in advance.

[jagabo]

Try single pass constant quality mode.

[Midzuki]

And if you find the time and the patience to,
play around with the damn quantize matrices ---
--- which are VERY different from the ones that
work well for MPEG-1/2 or DivX/XviD, BTW

[ronnylov]

I think there is some internal deblocking option you can use in x264, have you tried that? In general with mp4 formats it is common that flat areas may have macroblocks even at high bitrates and adding a little bit of noise may help to hide it, but here you said the source already was nosiy...

Anyway I would go with jagabo's advice, try single pass constant quality mode to get the needed bitrate for a given quality level.

I have to admit I have never encoded x264... I mostly use XviD and encode at constant quantizer of 3 and never thinks about what bitrate is needed, the quality is OK anyway and then i play it on my TviX 4000 player with internal hard drive. If I had the newer TviX 4100 that handles h.264 I would probably encode x264 a lot...

[SatStorm]

Hi to all!

The Deblocking option is by default "on". The only thing I changed on the default settings, was to set "interlace" to on.

CQ is not good for my needs here: The point is to convert 1 VHS tape (3h.10min) to H264 at a ~ 4.300Mb filesize so to burn it, as data, to a DVD - R, for archiving reasons. CBR is a great choices here, but it seems that a bitrate around 3100kb/s is not enough for my needs.

So the question that rised inside my head is: "What bitrate gives macroblock-free encodings on H264 for a PAL Full D1 under any occasion? ".

[Midzuki]

It's not helpful to increase the encoding bitrate when there is something that makes the encoder
use most of the given bitrate for "filesize padding" rather than for video compression.

[jagabo]

CQ is not good for my needs here: The point is to convert 1 VHS tape (3h.10min) to H264 at a ~ 4.300Mb filesize so to burn it, as data, to a DVD - R, for archiving reasons. CBR is a great choices here, but it seems that a bitrate around 3100kb/s is not enough for my needs.

Then you have no choice in bitrate, only CBR vs. VBR. VBR should be better. But if the source is very noise it may not help much.

So the question that rised inside my head is: "What bitrate gives macroblock-free encodings on H264 for a PAL Full D1 under any occasion? ".

Just as with any lossy DCT based codec, there is no bitrate that will ensure macroblock free encoding under all conditions.

[ronnylov]

Can't you do some tests with a part of the video that got blocky and make different encodings at different bitrates to see the limits? I mean since you are using CBR it does not matter if you only encode one minute or 3 hours if you just choose one of the diffucult sections of the video.

Maybe test some different quantizations matrices at your given bitrate can improve some also. I mean some matrices are optimized for low bitrate and other for high bitrate. Since this is quite a high bitrate maybe you get better results with a hig bitrate matrice, or maybe the bitrate is too low and you need a "low bitrate matrice". And why not use VBR? I think there are a lot of tweakable settings in x264. Some options probably gives better quality but may need more encoding time...

[SatStorm]

Well, testing is one option, but you have to have time for tests! My time is limited nowdays, too limited I have to say. But that's my problem, I'll find time...

The reason I asked the forum is if someone has done that before me, so not to have to invent the wheel from the start!

Regarding the H264 options, indeed they are plenty and in the case of x264, things change from time to time.
The Quantizations matrices help on lower bitrates I think, but I'll try some to see if that also occurs on higher bitrates.

Midzuki pointed something correct. Overkilling may be a problem in my case.

The other thing I was thinking, was virtualdub and vfm x264.

[poisondeathray]

SatStorm -

Your bitrate should be more than adequate, I doubt that is the problem

The x264 VfW version lacks several options that the CLI version possesses (including CQM's) and is updated less frequently.

If you are going to try the matrices out, I find that the easiest front end for the CLI version is avidemux which looks & behaves similarly to vdub. If you are short on time, you can just encode a 10sec clip with 'mark in'/'mark out' and readjust your settings and try again

Good luck!

[ronnylov]

One way to see "what bitrate is needed" is to make a compressibility test by using a single pass constant quality mode of a representative sample of your video. The bitrate you get is a hint of what bitrate you would need to get acceptable quality on the full video (if the resulting quality on the encoded test clip was OK).

[SatStorm]

My next attempt to encode to H264, is going to be this weekend. I'll try some things this time, to see if there going to be a difference.

I'll try avidemux this time.

[SatStorm]

OK, I did a convert a little earlier.

I read and change a couple of x264 settings and the first results look ok (far better my previous attempt for sure). I'll test further of course.