I-frames captures

[RaleighNC]

Can someone please explain i-frame captures. I have an aiw128/32 on a p3-550. Can I use i-frames to make svcd's.

TIA.

[jwhitlow]

I have an AIW Radeon. I use I-frame captures at 7.5K and then use TMPGEnc to encode to VCD.

[RaleighNC]

Why just vcd?? Are you capturing at 352x240@7.5??

[Darkbird]

There are three main types of video frames for mpeg. 'I' frames are like key-frames, they contain all of the information for that frame in that frame (similar to a jpeg compressed still image). 'B' frames contain info about the changes between two frames. 'P' frames contain predictive info about frame differences. B and P type frames are much smaller than I type frames, and that is where mpeg streams save the most bytes for high compression.

An 'I' frame only stream (or capture) can be of very high quality if the bit rate is high enough. But, it also ends up being very large in file size compared to a normal mpeg stream. You would NOT want to make a VCD or SVCD directly from an 'I' frame only stream. Instead people use 'I' frame only captures in order to have a high quality capture, then re-encode to a lower bitrate using the normal combination of 'I', 'B', and 'P' frames. Re-encoding with a software based encoder allows optimizing the video stream for quality with multiple passes. Today's consumer grade real-time hardware encoders do not do a very good job of optimizing bitstreams at low video rates. And today's computers also have a hard time encoding optimized streams at low bitrates in real-time.

[RaleighNC]

So you're saying I-frame capture for quality source file, and then re-encode to svcd spec.

Do I need to capture 720x480res and convert to 480x480, or capture already @ 480x480??

How about bitrate?? SVCD is 2.6. Should I capture at a higher bit > 4mbit?? or at 2.6mbit??

My whole goal is to make the best source mpeg-2 file that's < 4gb and that my cpu and hd can manage.

TIA Darkbird and anyone else that can help.

[RaleighNC]

bump

[fingernailX]

This may not be of help but the Matrox 2000 & 2500 (respected $1000 Mpeg2 capture cards) capture (hardware assisted) in I-Frame at full D1 720 x 480 at bitrates of 10 to 25 Mbps. see bottom of page http://www.matrox.com/videoweb/products/enduser/rt_specifications.htm
At the high end your are at DV bitrates. I would be very interested to see which compresses faster and better - a DV.avi or a I-Frame only 25 Mbps mpg. file. I now have a new mission over the coming weekend.

[Darkbird]

If you capture at only 2.6Mbps there is little reason for re-encoding to 2.6Mbps as you will just be losing quality.

Quick calc:
4000(MB) * 8 = 32000Mbits
90 minute movie * 60 = 5400 seconds
32000 / 5400 = 5.925Mbps
since 4GB is actually greater than 32000Mbits you could go to about 6Mbps for a 90 minute movie.

Now whether you can actually create an 'I' frame only stream at 720x480 and keep the data rate at 6Mbps or less is a good question. I guess the best way to find out is to try it.

The final encoding should be better when done from a 720x480 capture rather than 480x480 if you can manage that resolution (and bit rate). However, using 480x480 would probably be faster.

[Darkbird]

fingernailX: Your test will reveal which installed codec is faster since dv data rate is also 25Mbps. The winner may change with different codec versions from different vendors. Still, it will be interesting to see the results, if you wouldn't mind posting them for us. Give us the details of your setup, what codecs, from what vendor, etc.