DV (dvsd) To MPEG4 (H264 avc)?

[guada 2]

Hello everyone,

I achieved a lot of transfer DV video WMV9 to with Canopus Procoder 1.5.
And, I would be interested to know your opinion on 2 privileged methods: DV to MPEG4.
Of course, I speak in term of quality A/V and no in time of encoding.

- DV / Pinnacle 9.3.5 / huffyuv / VD + HS 264 /. Mp4muxer or
- DV / Nero Recode 3

Thank you.

[jggimi]

Moved to DV forum.

[guada 2]

Thank you jggimi, you were faster than me.

Sorry :)

[guada 2]

I think that I was mistaken of post unless my questions don't interest anyone. :(
I consider ungrounded then.

Sorry.

[bb]

Well, I do DV -> MPEG-2 most of the time, but I'd be interested in opinions on DV -> MPEG-4 AVC, too. I tried Nero Recode so far, but without any filtering the result was disappointing (as I expected it to be).

bb

[guada 2]

The format Mpeg 2, I use it solely with the DVDs and no with the DV format.
I would tell to you why.
Evidently, I don't have anything to envy my software of CCE encoding 2.70 Pro.

I noted that the capacity of the bitrate of CCE was sharpened so, that when the video camera moved, or changed direction suddenly. CCE showed signs of this report: a picture that curls..

It is for it I chose Procoder of Canopus to convert my DV in WMV9.
I use the Basic mode in order to exploit the WMV9 codec in Stream, that i reappoint with a big debit and the result is surprising, so much by fluidity and the speed.

With regard to my demand, I never made test with Nero H264 AVC.
Apparently, you don't seem to appreciate its potential.
I think that I am going to enact on my method.

Thank you for your commentary.
One would say a true "BB" baby." No it was just to laugh. :)

To soon bb ;)

[bb]

Originally posted by guada 2
[...]
Apparently, you don't seem to appreciate its potential.
[...]
No, Nero Digital is a very good codec, but for DV sources you need filtering in order to get good results. The same is true for DivX, MPEG-2, etc.

bb

[WorBry]

bb. I do a fair bit of encoding with Pal Type II DV sources (using the Sony DV software codec for decompression) and have recently turned my attention to Nero AVC (having upgraded from Nero 5 to 6). I've been using a simple AVS script for all MPEG4 and VP6.2 encodes and acheived pretty good results, especially with VP6.2

avisource("E:\Clip.avi")
ConvertToYUY2(interlaced=true)
ColorYUV(off_y=-16,gain_y=16) to correct for luma shift
KernelDeint(order=0,sharp=true)
LanczosResize(720,544)
Undot()

What filters would you recommend for optimum Nero AVC output? Would you recommend using the internal deinterlacer in Nero Recode instead?

Thanks

[bb]

@WorBry:
I don't believe that Nero's internal deinterlacer can compete with good AviSynth deinterlacers like KernelDeint, but I must admit that I have not compared deinterlacers for quite some time, because I usually record progressive DV.

Undot is a good filter, but if you need more quality for the bitrate, you should try temporal smoothing filters, too. My favourite is Convolution3D, which I often use in conjunction with low light shots with lots of noise. This kind of filtering is recommended for all MPEG variants, including MPEG-4 AVC.

bb

[guada 2]

"I don't believe Nero's that internal deinterlacer can compete with good AviSynth deinterlacers KernelDeint like, goal I must admitted I that haggard not compared deinterlacers heart quite some time, because I usually record progressive DV ".


BB do you thinks that the registration really makes itself in progressive mode, because I doubt between the Reinterlacing or the deinterlacing.
You can clarify my thought, because all this process sometimes challenge myself. :confused:

[WorBry]

Thanks bb.

The last time I tried Convolution3D I got a bit "convoluted" myself, but I'll have another go as a fair amount of my indoor home video footage is on the "lowish light" side (I must invest in a decent video light; the piddly 3Watt light I bought for my Sony DCR PC115E camcorder doesnt really do the trick).

I'm also doing some compares of Undot vs RemoveGrain with my PAL DV sources and from my initial tests (using EverWicked's Video Quality Studio),RemoveGrain definitely improves the PSNR of DivX, XviD and VP6.2 encodes. Any idea how to measure the PSNR of Nero AVC?

[bb]

Originally posted by guada 2
"I don't believe Nero's that internal deinterlacer can compete with good AviSynth deinterlacers KernelDeint like, goal I must admitted I that haggard not compared deinterlacers heart quite some time, because I usually record progressive DV ".
If you tried to quote me here, then the quotation is wrong... Are you using an online translation tool?

Originally posted by guada 2 BB do you thinks that the registration really makes itself in progressive mode, because I doubt between the Reinterlacing or the deinterlacing.
You can clarify my thought, because all this process sometimes challenge myself. :confused: [/B]
I'm afraid I don't understand what you are trying to say. I know, your native language is French, no problem.
Let me guess: If you wanted to ask whether my camcorder is really capable of shooting progressive video, the answer is yes: My Canon MV3 has a progressive scan option, thus I can go without the deinterlacing hassle. On the other hand interlaced video has its advantages, too. The temporal resolution is higher, so you get smoother movement on interlaced displays.

bb

[bb]

Originally posted by WorBry
Thanks bb.
Very welcome :)

Originally posted by WorBry
The last time I tried Convolution3D I got a bit "convoluted" myself, but I'll have another go as a fair amount of my indoor home video footage is on the "lowish light" side (I must invest in a decent video light; the piddly 3Watt light I bought for my Sony DCR PC115E camcorder doesnt really do the trick).
True: you need a good video light; these small video lamps - sometimes even built into the camcorder - aren't good for anything, if you ask me.

Originally posted by WorBry
I'm also doing some compares of Undot vs RemoveGrain with my PAL DV sources and from my initial tests (using EverWicked's Video Quality Studio),RemoveGrain definitely improves the PSNR of DivX, XviD and VP6.2 encodes. Any idea how to measure the PSNR of Nero AVC?
I'm afraid I have no idea how to measure the PSNR of Nero AVC - never tried it myself. Maybe Doom9 has a clue; if I remember correctly, he's done some PSNR comparisons.

bb

[WorBry]

bb

Thanks for the Convolution3D recommendation. It has markedly improved the compressibility of my Pal DV sources and proved an effective means of smoothing out the slight "scaliness" that was noticable (particularly on skin surfaces) with all MPEG4 encodes, and VP6 and H264 (AVC). So now I've modified my standard script to include Conv3D with a "light" setting:

avisource("Clip.avi")
ConvertToYUY2(interlaced=true)
ColorYUV(off_y=-16,gain_y=16)
KernelDeint(order=0,sharp=true)
LanczosResize(720,544)
Convolution3D(1,4,5,3,4,2.8,0)
RemoveGrain(mode=2)

BTW - I figured out how to measure PSNR of the Nero AVC encodes from Sagittaires Metric Test Challenge thread over in New AV Formats-Codecs:

http://forum.doom9.org/showthread.php?s=&threadid=90784

The results are very impressive. The only downside for me is the high CPU demand of AVC encodes compared to MPEG4 and MPEG2 (at least on my AMD XP2800, 1GDDR RAM). I appreciate that AVC is a complex codec, but would (dare to) suggest that decoder performance really needs improving if it is to receive wider acceptance as the new compression standard destined to supercede MPEG4 and DVD. I'll probably receive crtitique for that remark, but I think it's a fair comment at this point - not everyone has a fast PC.

[bb]

WorBry,

first of all thank you for posting your script.

AVC needs more processing power, that's for sure. But you may have noticed that Nero Digital capable standalones are already available, so I guess chances for the new codecs aren't as bad as you believe.

By the way: has anybody tried out the current version of AutoGK with DV sources? I haven't yet, but from the description it should be able to deal with AVIs now.

bb

[WorBry]

Hi bb,

Yes I've done a few initial tests with AutoGK v2.00b and 2.01b using PAL DV source files. AutoGK uses the DirectShow DV filter as decoder and in my experience achieves much lower compression than the Sony DV codec that I normally use. For example here are the results with one test clip:

1.AutoGK XviD 720x544, 75% quality (single pass) - resultant bitrate 6300kbps

2.With Sony DV codec, AVS script (as detailed in my previous post, but without Convolution3D), VirtualDubMod fast recompress, XviD (AS@L5 Single pass, Target Quant 2.67, MPEG, 1 B-frame, Motion Precision - Ultra-high Mode 1 Chroma motion, Quants max/min I 2/3,P 2/3, B 2/3, modifier - Chroma optimizer) - resultant bitrate 3200kbps (3000kbps with Convolution3D included).

Obviously quite a difference. Havent tried DivX yet.