PDA

View Full Version : Mainconcept DV vs Huffyuv codec


Mark Fredrickson
3rd January 2002, 04:25
I'm currently using Huffyuv to capture my home videos. I then encode them to DivX for permanent storage. I recently purchased a DV camcorder and have been playing with capturing using the Mainconcept DV codec.

I still have 10 years of VHS analog capture and read that I could capture the VHS videos using the Mainconcept DV codec (using my analog camcorder). I'm interested in doing this since Huffyuv has a very serious bug where it fails if it encounters blank portions of video tape (heavy static).

Could someone please shed some light on whether this is a good idea or not? Is the Mainconcept DV codec a lossless codec? Does it provide similar compression as Huffyuv?

Thanks,

Mark.

Guest
3rd January 2002, 05:09
I'm currently using Huffyuv to capture my home videos. I then encode them to DivX for permanent storage. I recently purchased a DV camcorder and have been playing with capturing using the Mainconcept DV codec.When capturing from a DV camcorder, the camcorder does the encoding and then just transfers the DV file to your computer via firewire. The Mainconcept codec only comes into play if you want to play the DV file on your computer, or decode it, possibly edit it, and re-encode it to DV (or to DivX).I still have 10 years of VHS analog capture and read that I could capture the VHS videos using the Mainconcept DV codec (using my analog camcorder). I'm interested in doing this since Huffyuv has a very serious bug where it fails if it encounters blank portions of video tape (heavy static).I've never heard of anyone doing this and doubt that the Mainconcept encoder is fast enough. What analog capture card would you use and what application? If it works, great. If not, then you could use the Dazzle Hollywood Bridge. But doesn't your DV camcorder have analog in?Could someone please shed some light on whether this is a good idea or not? Is the Mainconcept DV codec a lossless codec? Does it provide similar compression as Huffyuv?All DV codecs are lossy. They will compress better than Huffyuv. There are possible disadvantages to getting DV format, depending on your final target. For example, the pixel aspect ratio is not square, and there are some well-known artifacts in DV (blocking quilting, and mosquito noise). But your results using DV capture will be acceptable, especially considering that your source material is only VHS tape.

Mark Fredrickson
3rd January 2002, 06:36
Donald,

I got the idea from Markus Zingg's AVI_IO web site http://www.nct.ch/multimedia/avi_io/ where he says

"Using the mentioned DV codec from Mainconcept it's even possible to capture from analog sources into DV format in realtime using AVI_IO. You need a 700Mhz or faster processor for it and an analog capture device which is supporting the DV resolution for captures (most TV tuner boards will do). This is having the advantage, that older analog footage does not have to be rendered in the NLE application if it's used along with DV footage."

I've just finishing about 60 hours of 8mm video capture (8 hrs to go), then I'm off into my 10 years of VHS. It's been the 8mm videos that have given me a headache with Huffyuv. I traced it down to when I let the battery go dead while filming, it leaves a small gap of blank tape and Huffyuv really hates that. I want the best DivX output I can get so I want to capture with a lossless codec. Since you said that the Mainconcept DV codec is lossy, maybe I should stay with Huffyuv (unless my VHS tapes also have problems).

I didn't think about passing my analog video through my DV camera. Unless I will get much better quality I'm not sure if I want to put that much wear and tear on my new camera. What'dya think?

And again, thanks for the great filters.

Mark.

bb
3rd January 2002, 07:40
AFAIK, Huffyuv is the only lossless capture codec available. I don't use it because of the huge amounts of disk space needed. You might want to test the PicVideo MJPEG codec, which gives very good results at quality 19 (you could even try 20).

Capturing with a camcorder is not an option for me, because I'm limited to 90 minutes using a 60 min tape in long play mode. So what if I want to capture 220 min "Dances With The Wolves"?

I thought about using a DV codec for capturing, too. Didn't try yet, but I'll do this weekend. DV compresses 5:1, yields good results, and it's good for video editing. Let's see if it can compete with MJPEG...

bb

Mark Fredrickson
3rd January 2002, 08:07
bb,

It took me a while to figure out what AFAIK meant. That was pretty good. I tried PicVideo and couldn't get it to work very well for me but I didn't try too hard since I have enough disk space for a 2 hour Huffyuv (60 gb).

Huffyuv compresses at around 2.5:1, PicVideo at 8:1, so a DV codec looks like it's about in the middle. I may also give it a try later (I'm encoding right now).

Let me know how you get on, ok?

Thanks,

Mark.

Guest
3rd January 2002, 13:59
Originally posted by Mark Fredrickson
I got the idea from Markus Zingg's AVI_IO web site http://www.nct.ch/multimedia/avi_io/ where he says...

Very interesting!

I want the best DivX output I can get so I want to capture with a lossless codec. Since you said that the Mainconcept DV codec is lossy, maybe I should stay with Huffyuv (unless my VHS tapes also have problems).
Agreed.

I didn't think about passing my analog video through my DV camera. Unless I will get much better quality I'm not sure if I want to put that much wear and tear on my new camera. What'dya think?

Lots of people do it, although I understand your concern. You should seriously consider a device like the Dazzle Hollywood Bridge if you really want DV capture. But realize that if you capture DV, you'll have to do a resize to get back to square pixel aspect ratio for your DivX encodings. That will be a lengthy processing step and will slightly degrade quality. And the DV artifacts are a real concern. If I were you I'd stick with the HUFFYUV capture.

And again, thanks for the great filters.

Thanks. I'm planning to branch out into Avisynth filters.

BTW, the compression ration of Picvideo can be configured. DV is fixed at 5:1.

Cart
3rd January 2002, 22:27
In terms of file size, DV is about 3.6 MB/sec, although I suppose this could change a little, depending on how you capture the audio.

Anyways, I wanted to welcome you to the forum, Don. You might remember me from the string of annoying e-mails I sent you about Telecide and other filters. It's nice to see you here.

-Luke/Cart
http://www.geocities.com/lukesvideo/index.html

Mark Fredrickson
4th January 2002, 06:13
Sorry to be such a bother, but I'm new to this DV stuff. I've gotten pretty good at analog capture and encoding (thanks a bunch to you Luke) but there are still a few items I'm confused about with DV capture.

1. I have the Mainconcept DV codec installed and after I captured my DV video and then query the Avi properties, I notice that the 'Video Compression' says MC_DVD. Does this mean that I compressed whatever was on the DV tape to MC_DVD or does it mean that I just copied the DV tape and will be using the MC_DVD to decode it?

2. Don, you mention resizing back to square pixels. I don't quite understand the square pixels part (please excuse my ignorance). Could you maybe elaborate on that? Everything I've done has been at the NTSC standard of 4:3.

3. As far as the resizing goes, I do have a question about the advantages/disavantages to resizing from DV 720x480 to 640x480. All my analog videos are captured at 640x480 and I was wondering if I should resize my DV videos or just leave them at 720x480. Ultimately they will be played back on both the computer monitor and the TV via the computer's TV-Out (using Zoom Player). However I do want to try to maximize my compression ratio (with respect to quality being the highest factor).

Here are my specs:

Dual P3 1G, 512mb ram, 2 raided IBM 45gb Deskstars, ATI Radeon 64DDR video, Win2K Pro, EditStudio 3.3.

Average DivX video is 1gb/hr. This leaves me with about 50 min. per CDROM. This time next year, I'll probably be recording them onto DVD (only for storage, not for playback).

Thanks for all your help,

Mark.

PS. Hey Luke, so far EditStudio has given me perfect audio sync. I have a licensed copy of AVI_IO but I'm now using EditStudio because it's WDM complient (I think), has many more nice features, and usually recovers the capture after Huffyuv crashes. Check it out and let me know what you think, ok? Oh, and it's DV capture is absolutely superb!!! Yeah, and one more thing, the capture part is FREE!

Guest
4th January 2002, 13:50
@Mark

On pixel aspect ratio:
http://toolbox.sgi.com/TasteOfDT/documents/video/lurker/pixelaspect.html

You need to resize width by a factor of 10/11 and then crop as desired.

Guest
4th January 2002, 13:55
@Cart

Sure I remember you, Cart. I still have some mails from you I need to reply to! You were very helpful to me with the development of several of my filters and you corrected misconceptions I (too) had about aspect ratios and DV. Thanks to you!

Cart
4th January 2002, 22:02
Mark:

1) I believe that's just the name of the decoder.

2 & 3) You'll probably want to resize the video so it has square pixels because your monitor expects 4:3 video and your TV-out probably uses it too.

Also, I think there may be one (or more) other lossless codecs besides Huffyuv. I seem to to recall seeing a Japanese lossless codec. ("AzLib" maybe?) This page also has some possible links near the bottom:
http://www01.u-page.so-net.ne.jp/jc4/hoehoe/link.html
Remember, http://babel.altavista.com is your friend.

I'll see if I can check out EditStudio sometime.

-Cart
http://www.geocities.com/lukesvideo/index.html

Cart
4th January 2002, 22:13
Don,

You do realize that by posting here, you've opened a new avenue for filter request/complaints, right? You must enjoy pain ;)

Anyways, it is I who must thank you for providing such great filters. Your efforts are truly appreciated.

-Luke
http://www.geocities.com/lukesvideo/index.html

Mark Fredrickson
5th January 2002, 06:18
Hi guys,

Ok, I think I'm getting a little closer to being ready to encode my first DV. But ....

1. I understand how to resize from 720 to 640, that's pretty straightforward. But I'm completely lost with this square pixel thing. I read your recommended web page Don, and it was waaaaay over my head. Could you please give me some pointers as to how to get square pixels using VirtualDub?

2. When I resize down from 720 to 640, which mode do you recommend (bilinear, bicubic, or the precise versions)? I've read that bilinear is for downsizing while bicubic is for resizing up, but there seems to be a camp that favors bicubic for everything but they're mostly DVD rippers.

3. Don, those DV artifacts that you mentioned (blocking quilting, and mosquito noise). Are there any filters or any processing with VirtualDub I should be doing to eliminate or reduce them?

4. I have noticed on my DV recording an aura around my children when they are far away from the camera. They were in snow and I could see a shimmering effect that completely surrounded them. Is this one of those artifacts? Can I get rid of that with a filter?

5. Is recording my DV in progressive scan mode and encoding to DivX the same as recording my analog video, deinterlacing it, and then encoding to DivX? If I'm deinterlacing all my video before encoding (so I can watch it on my computer), is there any reason why I should be recording my DV in interlaced mode and then deinterlacing it when I can just record in progressive scan mode?

6. Sorry for all the questions. I've looked around for a DV Capture guide that would explain all this but had no luck. Did I miss one?

Thanks again and again,

Mark.

PS. Luke, I downloaded the LCL codec from that Japanese web site and I'll give it a go this weekend. Here is a bit from the readme file:

LCL is Loss-Less Codec for Video For Windows.
LCL is suitable for digital animation or the animation of 3DCGs.
LCL contains two kinds of codecs by the use.
AVIzlib: It is the codec of the rate of high compression which used zlib.
AVImszh: It is a codec for digital animation.
Kenji Oshima owns the copyright of this software.
Both of Jean-loup Gailly and Mark Adlert own zlib compression library.

Cart
5th January 2002, 10:41
1) The only other page I know might not provide too much help, but here you go:
http://www.uwasa.fi/~f76998/video/svcd/faq/

Here's my attempt at an explanation:
For most video, the pixels are meant to be viewed as square. This means that on a monitor, a square pixel 640x480 should take up 640x480 pixels. The pixels in DV video, on the other hand, are taller than they are wide. However, when viewing a DV video on a monitor (a square pixel device), are pixels are displayed as square. This forces the DV pixels, which are supposed to be horizontally thin, to stretch out so they are square. The result is that the video is wider than it should be. To correct for this, the video must be shrunk horizontally through resizing. Thus, you resize a 720x480 DV video to ~654x480 to make it look correct on a square pixel monitor.

To make it even more confusing, DV video has more horizontal overscan than a standard 4:3 video. For example, if you filmed a ruler running from the left side of the image to the right, a 4:3 video might show only 6 inches of the ruler whereas a DV video might show 7 inches. That's why a DV video resized for square pixels is ~654 pixels wide instead of 640. Cropping to 640 pixels makes it the same as a standard 4:3 video.

2) Bicubic resizing produces a slightly sharper image than bilinear resizing. Avery Lee says that using Precise Bicubic with A=-0.60 produces the most accurate image (i.e. the least amount of sharpening or blurring).

5) It depends on the camera. I've heard that the progessive mode on some models simply drops one of the fields. In this case, you would obtain additional detail by recording interlaced and using a filter such as, oh, let's say Smart Deinterlacer.

6) Maybe http://www.labdv.com ?

-Cart/Luke
http://www.geocities.com/lukesvideo/index.html

Guest
5th January 2002, 14:11
@Mark

1. You don't need to understand it! Just resize to 654 x 480 and then crop to 640 x 480 as Cart said. That's for NTSC of course. It's different for PAL.

2. Cart said bicubic is sharper than bilinear. That can be true but is not a complete answer. A proper bicubic will not look sharper; it will just be a more accurate interpolation and will not have the aliasing that bilinear introduces. As a side effect of using bicubic, you *can* choose parameters such that there is a sharpening effect. But you use bicubic because it gives you a more accurate resizing.

3. Post an URL to some sample frames and we can more properly make recommendations. Gentle application of temporal and spatial smoothers may help. But since your source is only VHS tape, the artifacting may not be triggered to an objectionable extent.

4. I'd need to see a sample frame. Likely that is mosquito noise due to the sharp transition to the snow. See 3 above.

5. See Cart's answer.

jj59
5th January 2002, 16:35
@Mark,
6) here are a couple of links to some sites which might be usefull but perhaps you already visited them

English
UT CK's site http://go.to/videocapture
VCDHelp http://www.vcdhelp.com/capture.htm
Ross McLennan http://geocities.com/aussie01au/index.html
DV to SVCD guide http://www.geocities.com/dvtosvcd/DVtoSVCD.htm

German
Stefan Uchrins http://www.uchrin.de/svcd/
DVD-SVCD forum http://www.dvd-svcd-forum.de/cgi-bin/dvd_board/ikonboard.cgi
MB1 (special DV-XSVCD) http://www.mb1.de.vu/
Ultimate Board http://pub6.ezboard.com/bultimate (with a linklist in capture forum)
DVDBoard http://dvdboard.wpf.de/index.php?

The links are mostly for producing SVCD but on Doom9's great site there are lots of guides for producing DivX.

jj

bb
5th January 2002, 17:29
Bad luck: capturing from my WinTV using a DV codec failed. I'm using the Panasonic DV codec, and it showed up in capture mode only when I switched to RGB24; it wasn't present in the YUY2 list...

Furthermore, the codec failed to initialize when starting the capture. Whatever I tried, I got an error message saying something like "wrong format". Same effect in VirtualDub and in AVI_IO.

Sorry. Has anybody had more luck with different codecs? DVSoft? MainConcept?

bb

Cart
6th January 2002, 01:01
Don:
2) Thanks, that's what I wanted to say. :)

bb:
Maybe you need to choose a YUV 4:1:1 or YUV 4:2:0 format, which is what DV uses. Did you try going through every available video format? Perhaps you could even try forcing a format with VirtualDub's custom capture options. This page might help:
http://www.webartz.com/fourcc/

Mark:
If you're going to crop the video, I think it would be better to first crop from 720x480 to 704x480 and then resize to 640x480. This should be slightly faster and would have a slightly more accurate aspect ratio.

-Cart
http://www.geocities.com/lukesvideo/index.html

nsdn
9th January 2002, 02:16
interesting thread.

Just wanted to say that when I archived my old analogue (familly video) tapes, I dicked around with heaps of filters and stuff....and Donald thanks for your filters....particularly "smart smoother".

I did end up using the main concept DV codec with a paid for version of AVI_IO successfully for all of the archiving. This in fact work SO well I ended up dumping all of my filtering and was able to pretty near reproduce the original tape quality....without blocking...even though the source quality was pretty poor (in fact it was really crap!!!), I found as soon as I added noise reduction or smart smoothing.....I lost an unacceptable amount of detail.....in order for the filtering to make any differance. I found that fiddling around with the capture card video levels (simply chroma, contrast and brightness) had far more of an impact on quality of captures and conversions than any filtering or noise reduction techniques.

Interestingly however this really only worked for analogue captures....using the DV codec....and resizing to 352*288...using a MPEG 2 (XSVCD) stream. I still have not been able to do an acceptable DV conversion to SVCD (from a DV source) at SVCD resolutions (480*576) without some blocking with fast movement....continues to frustrate me...(although using CCE and DVD bit rates it works great...need to wait till DVD-+R is more cost effective)

I am puzzled since the quality of the DV capture is so good, yet it seems impossible to encode it without the resulting blocks with movement. I have seen DVDs that look worse as a source yet they encode well.....only thing I can think of is that DVD may use some colour depth reduction prior to encoding? just a thought.

bb
10th January 2002, 07:01
@nsdn:
If you have problems with DV sources, I recommend the filtering I suggested in this thread:

http://rilanparty.com/vbb/showthread.php?s=&threadid=11567

:cool:

bb

Mark Fredrickson
10th January 2002, 07:25
Everybody:

I tried the MainConceptDV codec also and couldn't get it to work. It would not show up in VirtualDub or AVI_IO. In EditStudio and iuVCR, it failed. Maybe because it's the evaluation version and only works decoding?

I also tried the LCL codec but it was buggy and it's compression was pretty bad. I'm going to stay with Huffyuv for now. I got the bug fixed (see related topic) so I'm happy.

Luke:

That's a good website http://www.labdv.com. I've bookmarked it. But I think it would be better for all of us if you had the time to update your guides on your website to include all this DV stuff. A lot of people start off there and go back for refresher courses. Also, how about adding this to Doom9's guides and maybe Nicki's too?

Don:

For my DV encoding to DivX, I set VirtualDub's resize filter to crop to 704x480, then resize down to 640x480 using Precise Bilinear, no other filters. As far as my old eyes can tell, it's pretty good. I'll play around with Bicubic and see if I can make out any difference. I'm still trying to get a snapshot of those artifacts. I'll post them tomorrow night.

Thanks to all of you. I think I'm ready to dive into the DV world.

Mark.

Zhnujm
10th January 2002, 20:29
did you use a resolution of 720x576 or 720x480 ? only then the mainconcept codec shows up in virtualdub. the demo supports both decoding + coding. but its much slower than the mjpeg codec.

Mark Fredrickson
11th January 2002, 02:10
Hey Zhnujm, thanks for the tip! The MainConceptDV codec now shows up in both AVI_IO and VirtualDUb. I'll play around with it and report back.

tai
11th January 2002, 02:24
How do I resize/crop/pad if my DV resolution is 720x576 ? Thanks very much.

bb
11th January 2002, 18:55
Now I know why my setup didn't work: my Panasonic DV codec needs RGB color space AND 720x576 resolution!

Ok, here are my results.


Capture Test
============

System: Athlon Thunderbird 1.3 GHz, 80GB @ 5400/min,
WinTV PCI, ATI Radeon VE, 128MB

Video: 1000 frames, 720x576
Audio: none

DV: Panasonic
MJPEG: PicVideo, quality 19

Codec Type Size CPU Color Space
----------------------------------------------
DV Cartoon 139.672 KB 90% RGB24
MJPEG Cartoon 160.328 KB 60% RGB24
MJPEG Cartoon 140.772 KB 50% YUY2
HUFFYUV Cartoon 369.808 KB 70% YUY2
DV Movie 140.656 KB RGB24
MJPEG Movie 149.800 KB RGB24
HUFFYUV Movie 376.660 KB YUY2


The quality of the DV capture was pretty good, especially for the cartoons. Compared to MJPEG at quality 19 the size is even smaller. But the CPU usage was high - too high for capturing safely on my machine. I guess the Mainconcept codec might be a better choice.
BTW: I wasn't able to capture with HUFFYUV at 720x576 in RGB24 color space. Data rate was too high.

Conclusion: DV capturing is an alternative to MJPEG for fast computers. But you are limited to a fixed resolution (720x576 for PAL), and you have to use RGB color space.

bb

Damn: The table would really look better in a non-proportional font...
Just copy and paste it to your favourite editor.

Zhnujm
11th January 2002, 20:18
the mainconcept dv codec runs near 100% cpu usage on my p3-866 and it supports YUV mode.

Cart
14th January 2002, 01:07
Mark,

It would be kind of hard for me to make a DV guide, considering that my only experience with DV is some sample AVI I downloaded a while ago. Plus I've got a bunch of of evil classes this semester, so my time runs thin.

tai,

What kind of output resolution do you want? PAL DV has a pixel aspect ratio of 54/59 (height/width), so you have to resize based on that ratio to produce square pixels. If you missed the link, see:
http://www.uwasa.fi/~f76998/video/svcd/faq/

bb,
Thanks for the info. If possible, you could try using a YUV 4:2:0 format to see if the quality is the same while lowering the CPU usage. (PAL DV uses YUV 4:2:0.)

-Cart
http://www.geocities.com/lukesvideo/index.html

bb
14th January 2002, 06:41
In the meantime I tried an evaluation version of the Mainconcept codec (the one with the forced logo in the upper left corner - unfortunately I don't have the full version).

It uses significantly less CPU than my good old Panasonic DV codec: an average of about 60% on my machine. I'll also try the DVSoft codec, maybe this weekend.

BTW: I got ugly results with the DV codecs when trying to deinterlace. I'll have to do a lot more testclips to verify this phenomenon. Someone else got experience with capturing analogue TV using a DV codec?

Cart,
ok, I'll try, but probably not before this weekend. Sorry, I'm short of time, too.

bb