PIC Video MJPEG codec settings and interlacing

[Radu]

Hi all,

First, I would like to thank everybody for all the knowlwdge I have acquired just by reading the posts in this forum.

The Questions:

How do you use the "2 fields if more than nnn lines" in the Pic video codec ? Does that mean that if you check it the captured video will be interlaced ? If you uncheck it, does it mean that the video is progressive ?

I did a small test capturing with my Marvel G400 at 704x480 (NTSC) vith AVI_IO, one capture with the above option enabled, the other with the option diabled. The captures went OK.

When I examined the video in VirtualDub you can clearly see the interlacing artifacts in both clips. What is happening here ?

Suppose you have a video clip captured with Pic Video. The video is interlaced. If you need to apply a chain of filters(for example deinterlace, crop and resize) in VirtualDub, where in this chain does the deinterlacing filter go ? Is it the first one, does it have to work on the original video ? Or can it go at any stage ?

How do you decide when to do deinterlacing and when not to do it ?

(All of the above is based on a video clip captured from TV with either Pic Video or Matrox HW MJPEG)

Thank you for your feedback.
Radu

[Dr Mordrid]

If you select "2 fields if more...." then any captures in a resolution over the set height limit (typically 240 for NTSC and 288 for PAL) you get interlaced video. This is NORMAL since the standard is for full frame video in both formats to be interlaced.

You get interlacing artifacts in portions where there is significant horizontal motion between the two fields in a given frame. When this is a problem then deinterlacing the video using VirtualDUB's smart delacer plugin is in order. 'tis better to do this than to capture 240 high non-interlaced video because 240 throws away too much data that the smart delacer can use.

The food chain goes;

Capture interlaced. If in-frame motion causes lacing artifacts that cause problems then;

a. in VirtualDUB deinterlace the clip using the smart delacer plugin for VirtualDUB. This holds for using the clip later in a program like MPSro or Premiere as well.

Clip preparation before it hits the editor is a normal part of post production and VirtualDUB is a very valuable tool for this.

b. if encoding interlaced footage to MPEG in TMPGEnc and lace artifacts become an issue then apply the deinterlace filter with the "Even-Odd field (field, adaptation)" setting. Works great.

Dr. Mordrid

[Radu]

"I'm tellin' ya, they call him the Doctor for a reason. He seems to know everything that's going on. "

Quote from an old Movie.

Thank you Doc. Couldn't get a better answer. The artifacts were very visible on horizontal movement(like a comb on the side of the moving subject).

On a side note, I did not mention the rig I used for capture, I have an old" ABIT BP6 dualie with 2 Celerons oc'ed to 525 Mhz, 640 MB RAM. The capture drive is hooked to an Adaptec U2W SCSI
and is an old 50Gb Seagate (LVD). The video card is a Matrox Marvel G400. Runs W2K advanced server, as this machine has other duties to acomplish as well. (Domain controller for the home network, web server, some C++ development,...)

Sound , the excellent Turtle Beach. I got the card follwing the advice of the people posting in this forum.

I was using a SoundBlaster Live! MP3+ before (a better name is SystemBlaster! ) and I was getting lots of crashes(2 times to the point where I had to reinstall the OS from scratch) and a lot of dropped frames. Since I got the TB this machine can go for months without a glitch, and can capture MJPEG at 704x480(both Sw and HW) almost perfectly. (1-2 dropped frames in 10,000 frames, using VirtualDub or AVI_IO)

What I intend to do is capture some old VHS tapes I have and "The Simpsons" cartoons and store them in either DivX or MPEG2. Haven't decided yet.


Which of the 2 formats would be more suitable ?

I have tried DivX5 on a short clip in 2 pass encoding and the results (at least for me) were very good.
I still have to test TMPGenc.

[landrover]

Which of the 2 formats?

It all depends on your personal preferences. I too am very impressed by DivX5, and I never used the previous codecs in any serious way. For the past 2 years I've been using VCD created with TMPGenc, because of the widely recognised standard and the good quality; note that you can play VCD in any standalone DVD-player. Now that DivX is about to be accepted by the hardware manufacturing industry it certainly has the advantage of getting much more footage on 1 CD. I will probably use it in cases where I want to fit an entire movie on 1 CD, where until now I always used 2 CD's for one movie. I only hope will be compatible in the future, so that you can play version 4 or 5 encoded stuff when DivX10 is the standard.
Almost forgot to mention that there is another format that could be useful: Windows Media. That is, if you're not allergic to Micros*ck products.

landrover

[Dr Mordrid]

WMV will be an attractive option given that APEX is putting it in their DVD decks this fall.

It'll likely beat any MPEG-4/DivX format(s) into players by quite a time margin.

BTW: if you're encoding animations in TMPGEnc there are a couple of ther delace setting you might want to try;

"Even-Odd field (field, animation adaptation)"

and

"Even-Odd field (field, animation adaptation2)"

Try 'em and see on your particular footage and see how it looks.

Myself I prefer MPEG-2 for archival purposes, especially since my APEX decks can play plain MPEG-2 files burned to a plain ISO disk *today*.

MPEG-2 can also be converted back to a lossless (HuffYUV) or near-lossless (PICVideo @ quality 20) format easily for re-use. WMV, DivX etc. don't do this very cleanly in my experience.

Dr. Mordrid

[Radu]

Thank you Doc and landrover. The animation tips were great.

Just to summarize:

The original clip is 704x480 NTSC, interlaced, captured with HW Matrox MJPEG or PicVideo.

I have a standalone APEX AD-500W with enhaced firmware.

As a reference we will consider an original VHS tape playback quality on a good VHS VCR as acceptable to good.

Given these conditions,

What are the most common resize option values for the clip? (WxH)

How much footage (time wise) can you store on a 700MB CD-R using the size mentioned above, in VCD format ?

What is the quality compared to the referece ?

Please answer the same as above for SVCD and ISO MPEG-2 (mentioned by Doc)

Doc, it seems I have some homework to do. Can you please point me to some information about WMV ? Is this achieved using Windows media Encoder 8 ?

landrover, like everybody I strongly disagree with some of the practices used by Microsoft. But I am running their OS, so...

[landrover]

Hi Radu,

So am I, and sometimes I´m even happy with their products. Yes, WMV (Windows Media Video) is encoded with Windows Media Encoder. Some video editing applications can also encode to WMV, but I´ve never seen any of those programs giving reasonable results, so Media Encoder is the one to go for if you want to create good quality WMV. Since you can determine the bitrate yourself for WMV and DivX, it´s totally up to you how much footage you can fit on one CD. I managed to get a 95 minute rock concert on 1 CD in WMV7 at about 900Kbps, which resulted in reasonable quality. I have no personal experience with it, but many users claim that of all formats DivX gives the relatively highest quality at low bitrates.
Because VCD has a standardised constant bit rate (at 1150 Kbps) you can fit just about 70 minutes on a 700 MB CD. SVCD on the other hand, has variable bitrate, making it less predictable and more dependant on the nature of the footage how much you could fit on a CD. You should be able to get at least 35 minutes on a CD. Over the years, many users have created some none standard encoding templates for TMPGEnc, in some cases resulting in over 90 minutes of footage per CD (so I´ve read) but often compromising compatibility with software and hardware player applications.

landrover

[Flying dutchman]

One more remark about the "interlaced" setting in Picvideo:

If you only want to use the captured material for stuff like SVCD's, I most strongly advise to leave this setting UNCHECKED because the image quality is notably BETTER and SHARPER then. Checking or unchecking this option has no influence whatsoever on the rest of your workflow, full-res MJPG will still be interlaced no matter what.

Apart from slight differences in compression ratio, there's no really compelling reason to compress interlaced video into two fields rather than in a single frame.
In fact, two of the best Mpeg encoders (tMpegencoder and bbMpeg) compress interlaced SVCD video into whole frames rather than as separate fields, which you can verify using a tool like "teco bitrate".

This "interlaced" option in Picvideo was, as far as I know, only introduced to make the codec compatible to Zoran hardware MJPG decoders (such as in the Matrox Marvel). These Zoran chips expect the two fields in the frame to be compressed separately rather than as a single frame.
So you need to check this option only if you plan to output the captured file through the hardware MJPG decoder. (If I create a Picvideo AVI file, and tweak the file header a bit [change the fourCC into "dmb1"] my Marvel G200 happily decodes the MJPG in hardware and outputs it over the composite video connector...)