Capture Quality 1394 vs MJPEG

[Flying dutchman]

I bought a mini-DV camera last year (Canon MV30, a.k.a. Elura) to replace my old Metz (Panasonic) SVHS-C camcorder. I spent most of an afternoon in the photo shop and tried everything they sold.
Though my old camera gave a very acceptable sharp picture, the crispness and high contrast of the Canon made me hold my breath when I tested it in the shop. I could actually film off a newspaper and read it. In comparison to this camera, many others I tried (JVC, Sony, Panasonic, Hitachi) were bluntly disappointing.

What I'm trying to say, is that apart from the physical format (DV? Mini-DV? D8?) the quality of the OPTICS is essential for a good camera. Canon is a manufacturer of optical products; they mainly make binoculars and cameras and are darn good at non-glare coatings etc. They only started manufacturing stuff like cheapo computer printers much later.

The other brands I mentioned mainly make consumer electronics; they are good at VCR's, but optical lenses are a wholly different game. Many just don't have that much expertise and experience (yet). Sony, for example, buy the lenses for their more expensive cameras from Zeiss; apparently even they acknowledge that "the experts" make better lenses.

Maybe DV has some inherent disadvantages over hi-8 or SVHS. In my case, these are more than offset by the excellent optics of the camera. For a fair comparison of analog and digital formats, the rest of the hardware should be kept equal. Sony, for example, manufactures Hi-8 and D-8 cameras based on the same bodies and lenses. Has anybody ever compared the output of these lookalikes?

[Frank Marshall]

DVD and Hi8 vs DV

Thank you Doc for defending your position. It seems that Adam's experience with DV-to-MPEG encoding is indeed specific to DVD authoring.

"I'm sure that like many pros he's mainly seen DV encoded to 720x480 MPEG-2 for DVD at moderately high bitrates. With a large frame size works because most errors will be small enough in proportion to the whole frame to not cause many issues, but that's what typical users are doing is it? Nope...."

DVD encoding isn't what typical users are doing *now* but the first wave of prosumer-priced DVD recorders are arriving now and they'll just get cheaper from here on in. DVD authoring is where we're going, or at least that's what I'm aiming for.

So, I can accept that there's no reason to go digital for chromakeying or even for encoding to VCD, or even SVCD. I understand that DV is probably better for regular or 'natural' video work destined for tape or display on TV. However, considering what you said above regarding DVD encoding, there still remains the question of whether DV is in fact a superior format for encoding to DVD (using 'natural' video of course) or can I confidently stick with Hi8 for this application too?

Thank you.

[mrtj1]

my father has just bought a sony digi8 camera (dv was too expensive). apparently the tapes are the same size as his old 8mm tapes. what is the quality diff between this and dv, and between hi8 and digi8 tapes?

[Dr Mordrid]

For DVD encoding, presuming the absense of things that will trip up DV regardless of target format, DV should be OK 98% of the time.

The problems that will remain are as stated above: dark and solid color regions, fast moving subjects, diagonal lines inframe (trellises, fences at an angle, tile roofs at an angle etc.), keying and so on. Those will get the Hi8/YUV treatment.

Dr. Mordrid

[Vidgal]

Well, I'm more confused than ever. Seems like Doc's list of problems is longer than the list of advantages. I guess both analog and digital have there place, but there am I? Even if I could afford to keep both my Hi8 and the DV I have coming, I certainly couldn't be hauling around both camcorders. I do mostly personal stuff for myself, friends and family. However, I am trying to figure out a way to transmutate my business/human resources experience into doing human resources training and/or communicate video for businesses. Some of this would be video, some intranet, but I guess none could stay in native DV format. My new DV hasn't arrived, so it hasn't been opened. Perhaps I should just send it back and spend my money on the new DV burner?

[Dr Mordrid]

There can be a middle ground, if you're careful about the lighting of your shots (even & bright, avoiding shadows if possible) and don't get into keying (bluescreen etc.). Call it a bit of a hack.

The trick is to capture analog from the S-Video output of your DV cam. The footage will still have been encoded to DV, but the S-Video's 4:2:2 output will soften the stream a bit. You can then capture it with an analog board capable of YUV capture. This can then be encoded to MPEG with a bit less probability of solid color/shadow & concatenation artifacts than pure DV footage.

One might think that applying a soften filter to the MPEG during encoding would produce similar results, but it wouldn't. The analog capture will still look better because the effect is more subtle.

This is still not as good as going pure YUV in the situations that are sensitive to DV's limitations, but it can work.

The rub is that it won't get rid of the artifacts produced by diagonal lines, fast motion etc.

Dr. Mordrid

[Vidgal]

Thanks, Doc, that makes me feel a little bit better. The G400TV lives on. I was planning on leaving it in place anyway. I currently capture in either PicVideo or Huffy -- although I now seem to be getting the interlace lines on all captures. Before, it was only on Huffy. Guess I will try the uninstall / reinstall thing -- or wait for the new, improved tools?? When you do the un/re thing, should you also uninstall and reinstall the YUY crack and Huffy? Install order seems to be paramount with video stuff. Thanks again.

[Dr Mordrid]

A point of interest from another forum.

This was posted today by a member of MUG, the MediaStudio Users Group. This was sent to him by NewTek tech support in regards to DV artifacting in some of their software.

NewTek makes the Video Toaster, a pioneering video editing system that brought NLE to the desktop on the Amiga almost 15 years ago;

"xxxxxxxx,

You have some very legitimate points here, while some of these are problems with Aura (NOTE: Aura=NewTeks Video Paint type program), be aware that what you are trying to do also has some problems. Firstly DV has some serious multi-generational loss issues:

<A HREF="http://www.adamwilt.com/pix-codecs.html">http://www.adamwilt.com/pix-codecs.html</A>

<A HREF="http://www.adamwilt.com/pix-stress.html">http://www.adamwilt.com/pix-stress.html</A>

It appears that some codecs perform much better than others, so you might want to ensure that you are using one of the better codecs.I would also play with other codecs besides DV since DV on anime style graphics is not a very good fit since it is based on 8x8 kernels (NOTE: aka "macroblocks") which will give similar artifacts as JPEG compression.

== snipped a portion having to do with anime ==

Finally, a very good compromise would be to work in RTV files. These are uncompressed YUV frames, which are half the size of uncompressed 32bit RGB sequences. (NOTE: using RTV's would be very much the same as using HuffYUV captues)

In addition RTVs generally are read and written of disk a lot faster than AVIs. So, 1 second of uncompressed 32bit AVI takes 40Mb while a 1 second RTV clip would only be 20Mb. You would be very hard pushed to see the difference between these clips (and YUV422 is the native color space of 99.9% of all video anyhow.)

I hope that this helps, if it does not ... shoot me an email.

----- 2 -----

One way that might help minimize the very annoying 8x8 block problem of DV is to blur the image slightly before you re-encode it.

This will help the DV encoder concentrate in the lower-frequency areas of the image. Your problem is actually that the codec is trying to store as much high frequency information as possible, at the expense of some of the other components.

By blurring the image very slightly (especially if you have very sharp edges) you will help "guide" to codec to do what you want.To give another illustration. If you use JPEG to encode a hard page of white-on-black text it will introduce a lot of visible blocks in the result. However, if you first blur that image it will look a lot better when encoded.

Andrew"

Hmmm.....sounds familiar

My only criticism is that most blur filters can't be adjusted low enough to do it right. This is why I suggested capturing S-Video from a DV cam instead.

Dr. Mordrid

[Flying dutchman]

@mrtj1:
Hi-8 uses the very same tapes as D-8. That's the reason Sony invented D-8 in the first place; they are one of the world's largest manufacturers of tapes and hold the patents on 8 mm tapes! Plus the fact, that they can produce the camera's rather cheaply because they can use most of the hi-8 design...


@Doc:
What's this fast-movement stuff you're talking about? As far as I know, DV only does intra-frame DCT compression, not inter-frame, so what evil could motion do?

Furthermore, to eliminate the discrete-cosine artefacts that mess up a subsequent mpeg encoding, it could theoretically help to simply shift the frames 4 pixels diagonally. That way, the new MPEG macroblocks would overlap 4 DV macroblocks and "smooth them out" without the need for filters. If I had a firewire card, I would try this out right now, but unfortunately I don't...

[dchip]

I have a Canon Elura Mini DV and for some time I have noticed that my compressed DV footage lacks something. Things are improving since I learned to work with MPEG-2 and SVCD which is closer to what I actually shot than I have previously acheived. Still, overall, DV has not rocked my world the way I thought it would a year ago when I jumped in.

Don't get me wrong, when I finish my editing and play my timeline back onto DV tape, the quality is far better than anologue could ever touch, however (there's always a "however" in NLE) when heavy compression (and trying to do MPEG this-or-that) enters the picture the digital video quality quickly gives up its crisp picture. I don't regret going DV because I think what I am heading towards is DVD and when I get the hardware and software to make discs with those bitrates a lot of that artifacting should disappear. At least I hope so.

I just want to throw in one thing I noticed with DV camera when I was casually shooting a wedding I went to a few weeks ago. The church had very soft lighting but the altar had a large glass window about 20 feet behind it with bright sunlight entering. I was forced to try to film the exchange of vows against that.

The optics on the Canon Elura fought a fine fight and seemed to get me some good pictures. When I looked at the film later I noticed that the pictures, although decent, looked rather wierd. It almost looked as though the bride, groom and pastor had been pasted into the scene, like they were lifelike cardboard cutouts or something. The edges of them were very odd-looking. This effect was even worse when I compressed to MPEG-2. I know the lighting was the main problem but how much of this problem might have to do with the way DV uses light?

Finally, for those who haven't gone DV and are considering it, I would say if you have a nice analogue setup now, keep getting mileage out of it. If DVD content is what you are looking for, then perhaps pursue a good 3CCD Mini DV unit. Also, if you hear rumblings of future consumer mini DV cameras or DV editing packages being capable of acheiving that good colorspace Doc mentioned, I think it would definately be worth the wait for that.

[Ted222]

Final Decision

Well, I thought I would post my final conclusion after starting this thread.

I capture with the S-Video Input off both my Canon Optura PI Mini DV and my Sony TRV 99 HI8, using PICVIDEO. The results are not stunning, but are the best I have been able to achieve so far.

As a note of interest for those contemplating Camcorder purchases - the Canon's video captures appear to be a little bit better than the Sony's(clearer and brighter). And as Doc had mentioned, the Sony does better on Low Light situations. So, I have kept the old Sony and plan to use both camcorders.

I have 1394 cards for both my notebook (Dell 7500) and my desktop (Dell 4100) which I have yet to put to good use...I will get a Pioneer DVD recorder as soon as the prices drop...then maybe get some use out of them. Although for the notebook the card is fantastic.

thanks to everyone for all the discussion, its has really helped me to get set up,

Ted

[Dr Mordrid]

I'm not completely sure why DV has a problem with motion, but it does. I've got an idea or three, but more experimenting has to be done to really lock it down.

My current speculation is that it's an offshoot of the diagonal line stair-stepping that's such a problem with DV. I've come to call these #$%@ things "angle-facts".

Things like race cars (which have angled bodies), or even non-boxy cars going any decent speed, have generated what looks like angle-facts at their edges in some of my shots. Their motion just seems to makes them more obvious.

I think this is because the objects motion drags the angle-fact across adjacent macroblocks in the frame, enlarging the size of the original defect and giving it motion. This of course makes it more easily seen. I'm sure bright colors (or white) and high contrast edges make matters worse.

The problem with trying to prevent them is that not many moving objects consist completely of horizontal or vertical lines, which don't generate the angle-facts. Ohwell....

Example;

I did some shots during the Michigan 500 at Michigan Intl. Speedway last summer that really went sour on me. My wife was shooting with the Hi8 and in the footage where we overlapped our shots the difference was quite obvious. Painful in fact.

For the race fans: Juan Montoya beat Michael Andretti by .040 seconds after 162 lead changes (!!) The record qualifying speed was also set last year at 234.949 mph by Paul Tracy. MIS's 2 mile tri-oval makes for some great racing. Sometimes 3 cars will go abrest into a turn and come out the same way. Sheesh....

ANYHOW....the Hi8 footage exported by S-Viodeo and captured to HuffYUV looked fine.

Dr. Mordrid