Am I bit starving my video?

[magillagorilla]

I'm moving along with my VHS video restoration. The footage is smart bobed (QTGMC), so I get 59.94fps. My final output is Main Concept AVC in an MP4 container, 720x480, 59.94p, square pixel, CBR 14mbps, 2Pass. I don't remeber the GOP settings, I left them default in VMS.

The reason for Main Concept AVC MP4 is that it is native to VMS and is also a very common container.

Most of the footage I am working on I shot when I was young and had no concept of keeping the camera stable. So there is a lot of motion. I know on stable footage 5-8mbps AVC looks ok for VHS at 29.97fps. But because I have doubled my frames and considering how noisey and shakey the footage is, is 14mpbs enough?

It looks blockey in short spirts during fast motion but the same areas in the HUFFYUV footage are very murky and grainy. So it may just be the nature of the AVC codec rather than bit starving. I want quality but economy is still a factor. I could throw 40mbps at it but that would be silly. I need to find the sweet spot.

Opinions would be great.

[2Bdecided]

2 pass CBR?!

You should use constant quality VBR.

You shouldn't really be seeing blockiness at that bitrate, though it depends how you're watching it (e.g. 40" LCD with sharpness turned right up will make any SD look blocky!), how noisy the original is, and how good the encoder is.

Cheers,
David.

[jman98]

2 pass CBR?!

You should use constant quality VBR.

Let's be fair here. With CBR multi-pass is useless so 1 pass CBR is as good as 2 pass CBR so I understand that unstated point you are trying to make. With DVD resolutions like this a bit rate of 14 Mbps (that's 14000 Kbps) AVC is gigantic. This far exceeds DVD quality. I suppose the original poster could always test a higher bit rate and see what he thinks but my feeling is that throwing more bits at the source video or going to constant quality VBR isn't going to make it look any better than what he's got now with CBR.

[magillagorilla]

Thanks, I know 2 pass is kinda silly on CBR, but thats not my issue. My monitor is a 24in 1920x1080 Acer, the model number eludes me right now, I'm not in front of it. I know SD breaks up when you blow it up, I did consider that.

"This far exceeds DVD quality"

Yes but DVD does not run 59.94fps. I don't intend on returning the footage to an interlaced format. So to that example would 7mbps be enough for SD DVD quality. I guess my logic is 2x the frames 2x the data. Am I wrong?

[jagabo]

If you want to be sure your video isn't bitrate starved anywhere use constant quality encoding like 2Bdecided recommended.

[2Bdecided]

btw, I don't even know what VMS is, but most people use x264 via MeGUI for encoding + muxing to mp4. It takes the output of AVIsynth scripts directly as input.

Top bitrate for DVD is ~9Mbps MPEG-2.
MPEG-4 AVC (H.264) is 50%-100% more efficient than MPEG-2.
60p requires 50-100% more bitrate than 60i (30i).
So yes, 14Mbps AVC should typically be better quality than DVD.

I find interlaced home movies (especially very sharp ones with lots of movement) can be impossible to encode transparently for DVD. I have no problems with 50p x264 at a reasonable constant quality. It may peak above 14Mbps sometimes though with better quality settings.

FWIW at truly visually lossless settings (so you can pause it, and sharpen it, and still not see the difference from lossless), MPEG-1 (!) takes about 60-80Mbps for 704x576p50 deinterlaced from noisy home movies. I'm sure it's near-as-damn-it visually lossless at half that with that ancient codec though.


Long story short: with constant quality encoding (and QTGMC) you should be able to to a near-perfect job - though watch out for cartoonish / oil-painting-like denoising from QTGMC on some content.

Cheers,
David.

[magillagorilla]

QTGMC seems to be doing quite a good job on the footage, though it can be a bit soft at times it looks better than any other smartbobber I've tried. I don't pretend to understand all the settings for QTGMC, I just go QTGMC(Slow, Sharpness=x), I may fiddle with the sharpness value a bit. In fact, on some footage I still need to get some bad grain out even after QTGMC and run a light Neat Video on the chunky bits. I go VERY easy on NeatVideo, it will turn things plastic in a hurry, but it's a powerful tool when callibrated properly.

I was going on the same theory for bits. If the less effecient MPEG2 maxed out at 9mbps does OK on SD video at 29.97i then 14mbps for h.264 59.94p should work. So I am using a CBR 14mbps h.264 with MP4 wrapper.

VMS = Vegas Movie Studio

I have the regular consumer version 10 for dumb-dumbs like me. I am doing all my final color corrections here as the secondary color corrector is extremely effective on my footage. So, I don't really want to dump it to AVI again then encode it with HandBRake or MeGUI when VMS already has MainConcept's h.264 baked in.

So....
AVISynth does these tricks where needed: ChubbyRain2, fix chroma bleeding, chroma shift, 3Dcomb, QTGMC, colorYUV corrections, audio normalize

to VirtualDUB for: further chroma noise correction, Neat Video (helps with low level noise and dot crawl form the deinterlaceing)

to VMS for light sharpening, Secondary Color, Color, Boarders, Titles, Audio compressor limiter > MainConcept h.264

[jagabo]

Encode your problem sections with a lossless codec. Compare the lossless encode with the h.264 encode. That will tell you whether the loss of detail is happening in the filtering or the h.264 encoding.

[2Bdecided]

@magillagorilla, I admire your patience!

I don't know how Mainconcept compares with x264, but IIRC there's a way of frameserving from Vegas. I think it's more than a one-click solution, but it would let you encode using x264 without creating an intermediate file. Whether it's worth the trouble, I don't know, but I mention it since you seem to be going to a lot of trouble already!

And anyone who realises that NeatVideo has to be used very gently is probably doing the job properly.

Cheers,
David.

P.S. What's 3Dcomb? If you have first generation VHS home movies, then capturing via S-Video should avoid all rainbows. Composite Dubs can need serious derainbowing though.

[magillagorilla]

"What's 3Dcomb?" FFT3dfilter to be more specific. Its a 3D comb filter. It helps when I have some very noisy footage.

I have tried to frame serve Vegas using the debugmode frame server without much luck. I do dig x264 it looks great and is more effecient. The MainConcept h.264 looks about as good but cost me some extra file size. Honestly, my eye is not keen enough to say which is better looking; MainConsept h.264 or MeGUI x264. It seems like x264 is smoother at times, I don't know, maybe better temporaly, this could be due to adjustments in GOP. I'd rather not "F" with it.

" If you have first generation VHS home movies, then capturing via S-Video should avoid all rainbows." Right you are. Although my caps were S-Video the tapes have suffered ?-GEN transfers over composite. There is rainbows, chroma errors/noise galore. The secondary color corrector in VMS is allowing me to replace color in damaged color channels. This is tricky business and I'm sure I'm getting noticably amature results. But I can say without question anything I have done so far is an improvement over the original.

I can't wait to get in to my 1st gen SVHS VHS-C footage, it's going to seem like a cakewalk compared to this old VHS footage.

[2Bdecided]

I can't wait to get in to my 1st gen SVHS VHS-C footage, it's going to seem like a cakewalk compared to this old VHS footage.

Yes! Though myself I was a little disappointed - I could remember that on my CRT in 1992, first generation S-VHS sometimes looked better than broadcast, and clearly better than a VHS dub. S-VHS seemed like 120% of broadcast quality, with VHS being ~60% of broadcast quality.

Whereas in 2011, it all looks so much worse than HD that the first-generation S-VHS "advantage" doesn't seem so great. If 1080p50 is 100% quality, HDV is 90%, DV is 45%, S-VHS is 40%, VHS is 30%, and 2nd gen VHS is 20-25%.

At least the "wow" of seeing how people looked 20+ years ago distracts from the lousy quality for a while.

Cheers,
David.

[magillagorilla]

Ya, I was thinking more along the lines of ease of restoration of the 1st gen footage. Less fiddling, fewer filters etc. There is no comparing any analog footage to my 1080p HD stuff. But to the same parallel, the 1080 HD footage will look silly in 25 years when whatever new display format/camera puts it to shame. This is one reason I decided to deinterlace all my SD stuff. I don't see the point of returning any of this back to a SD format. I would also wager that in 25 years or less, interlaceing will no longer be used.

Some of this VHS I have is so damaged, I tend to shoot for realistic skin tones, even if it blows out color in other areas of the footage. It's all wrecked anyway, I may as well focus on the people, which was the point of shooting most of it anyway.

I think I have settled for 14mbps CBR h.264. It looks very faithful to the HUFFYUV version. 14mbps may be a bit of overkill but it's still a fraction of the size of lossless and looks several grades better than MPEG2 at a similar bit rate. Also most of the footage is handheld and very noisy. The extra bits per second only help.

Then there is storage, you can get a 3TB drive for about $100. So in another 10 years I think the size of these files will be a non-problem. I am also keeping copies of my original HUFFYUV caps on some old IDE drives. I may return to do this all over again if technology improves. The important thing is to get the stuff off the tapes. Good luck finding a working VCR in 25 years.