Capture I-Frames Only?

[DivadX]

I am using an ATI AIW 9000 with ATI MMC 9.1 to capture CAV LaserDiscs.

Does anybody know about capturing I-Frames only?

[BJ_M]

it is the same basicly as capturing to mjpeg ....

you need a high bit rate to do this ...

[lordsmurf]

There's no reason for I-frame only.
Not unless you're expecting to edit. Like BJ_M said, that's essentially MJPEG.

[BJ_M]

even the sony IMX mpeg is not all I frame and it IS for editing ...

it is IPPP

[DivadX]

Thanks for the input guys...I guess my real question is what is the relationship between I-P-B frames and overall video quality? Let's say you're doing 704X480 at 6-8mbps. I was assuming that basically the more I Frames the better? Or for example, giving everything else equal, would there be a video quality difference between 1-4-2 and 1-2-2 (I-P-B)? That is what I'm really curious about...

[lordsmurf]

JPEG, MJPEG and MPEG are similar, but please do not confuse them. A JPEG image takes like information, and squeezes it together in the file, compressing the file size. A Motion JPEG (known as MJPEG) does the same, at a frame-by-frame level. Each frame is analyzed for similarities, compressed, then the process is started over at the next frame. MPEG does this, as well as incorporates temporal compression using groups of pictures (the GOP). An MPEG is made up with I-frames, P-frames, and B-frames. The I-frame is the reference mark, and the P-frames and B-frames record only data that is difference from the I-frames. The intricacies of the actual encoding are far more complex than this, but this gives you a basic idea of how an MPEG works.

Quality is not dictates by this as much as it is by bitrate allocation.
Adequate bitrate on an I-P-B encode should be fine for you.

[iantri]

To elaborate on lordsmurfs post, at a low bitrate I-frame only would have a detrimental effect, and at a high rate, there would be very little difference, likely.

I-frames contain the complete picture information. P-frames contain only information that differs from the frame before it. B-frames contain only information that differs from the frame before it and is not in the frame following it.

At a low bitrate, there is simply not enough room to store a series of high-quality I-frames -- but when P and B frames are used, less bits are necessary for those and be spent on the I-frames instead for better picture quality.

[DivadX]

...giving everything else equal, would there be a video quality difference between 1-4-2 and 1-2-2 (I-P-B)? That is what I'm really curious about...

[BJ_M]

...giving everything else equal, would there be a video quality difference between 1-4-2 and 1-2-2 (I-P-B)? That is what I'm really curious about...

depends

[lordsmurf]

I'm actually VERY curious where BJ_M is going with "depends" (come on big fella, I'm all ears!)

Anyway, my experience shows that this depends largely on the software/hardware, in terms of IPB radius and the quality it gives out.

[DivadX]

Yes...maybe it depends on the price of tea in china...or if it is a full moon...

Guess I'm not being very clear...

"Giving everything equal" means same hardware, same software, same source, same resolution, same bitrate, etc...

Then if you captured it once at 1-4-2 and again at 1-2-2 would there be any video quality difference?

[lordsmurf]

It depends. Sometimes worse, sometime better. Lots of factors.

[petar]

...giving everything else equal, would there be a video quality difference between 1-4-2 and 1-2-2 (I-P-B)? That is what I'm really curious about...

Depends on the bitrate, on the video, on the capability of the encoder...

Shorter the GOP, more bitrate you need (because you use more I pictures in place of P or B). If spare bitrate you have, you may benefit from shortening the GOP.

Also, less complex the video (in terms of MPEG compression), less difference it will make (hence I said "may benefit").

But all this pretty much depends on the encoder and how it chooses to allocate the bitrate among the pictures and how it chooses to encode the P and B pictures' macroblocks.

So, we're back to what BJ_M and lordsmurf said, depends... Try and see.

[BJ_M]

even if a lot of high freq. content (trees , grass, smoke , water, brightly lite outdoors scenes) vs. lower freq content (animation, space , static objects , black, indoors low lighting) can change results ....


petar has a good answer

[fourthquark]

Isn't DV-format video (like used by MiniDV camcorders and output from ADVC-100) basically I-frame only?

[BJ_M]

yes - but it is not using jpeg or lzw type compression .

[DivadX]

OK I did some experiments and thought I would share the results:

I captured Side 1 of SW Definitive Collection CAV LaserDisc which contains a really good mix of lettering, space, action, lo-motion, and bright scenes using...

*ATI AIW 9000
*ATI MMC 9.1
*704X480
*6-8mbps
*LPCM audio

I did this 3 times with open GOP's 1-4-2, 1-2-2, and 1-1-1 (I-P-B).

Result #1 (1-4-2)
637MB file size, peak bit rate 8.3, avg bit rate 3.2

Result #2 (1-2-2)
667MB file size, peak bit rate 8.6, avg bit rate 3.3

Result #3 (1-1-1)
781MB file size, peak bit rate 8.2, avg bit rate 3.9

After authoring and watching on a TV they all look pretty good, but IMHO the "shorter" the GOP the better the video quality in my scenario.

[lordsmurf]

With active capturing, I find too many "P" frames can make quality degrade.

[MpegEncoder]

It's amazing how confusing this all can be.

I-frames are VERY expensive on the bandwidth (i.e., bitrate). The main reason that P and B frames are used is to limit the data to effectively get "free bits". An mpeg file with all I-frames would be a tremendous bitrate hog.

If you want to do editing, then you don't want mpeg to start with. Use something else and then convert to mpeg later (if that's what you need for your final output).