Questions on I frames in MPEG-2

tfc wrote on 12/7/2004, 11:17 PM
Several questions regarding I frames. Correct me if I'm wrong, but next to the bit rate in the MPEG-2 encoding process, I frames will be the next most important variable when it comes to producing high quality video. As I understand it, I frames are complete frames, right? Therefore, I would assume that the more complete I frames one has, the higher the quality of the final video. This brings up the next question. In the default setting on the Main Concept encoder for I frames, 15 is used. I believe this means that there are 15 frames between I frames. In the drop-down selection, it allows you to use anywhere from a value of 3 to 30, increasing at intervals of 3. I tried burning a short test sample at 3 and one at 30. The one at 30 was deemed non-compliant by DVD-A (2.0) and would have been re-encoded (yuck!). The one at 3 was accepted as compliant by DVD-A and burned successfully. It also played successfully on my set-top DVD player. Does anyone know the actual value or range of I frames that are deemed compliant by DVD-A? Even though the one I burned at 3 played on MY set-top, would it be compatible with other set-top's? Am I treading into the dangerous area of non-compatability with my tinkering of I frame values. Can anyone shed any light into exactly what is compliant in regards to I frames, with DVD-A (in other words, has the green check mark), and what would be universally compliant with all set-top DVD players?

Also, another question - what exactly defines a GOP (group of pictures)? Is it necessarily the number of frames between I frames, or is it something more specific? I don't recall any way to change a GOP in the Main Concept encoder in Vegas, is there a way?

Thanks!

Comments

Chienworks wrote on 12/8/2004, 3:59 AM
One problem with your premise: since I frames are complete pictures, they take more bits to store them. So, for a given bit rate, having more I frames means that the bits will be stretched thinner and there will be effectively more compression. More I frames therefore means less quality, not more. The whole reason for incomplete frames is to save bits and use them more wisely.

The advantage of more I frames is that you'll have less motion artifacts, but you'll have to use a higher bitrate to compensate for the quality loss.
farss wrote on 12/8/2004, 4:48 AM
The biggest single factor affecting any form of interframe compression system is the quality of the input material. As it gets worse the quality of the output gets exponentially worse. I've had half the players that tried to play one of my DVDs choke at the same point. The problem was when the lights went down, camera gain went up and all we had were black frames filled with noise. On the players that did cope all you see is noise frozen for the duration of the GOP, it looks terrible. Fix the noise and everything looked and worked great.
That's why the Hollywood DVDs look so good even at a very low bitrates, their material firstly is at least 4:2:2 and secondly no noise and it's pretty unusual for 35mm cameras to 'hose down' a shot.
Bob.
ScottW wrote on 12/8/2004, 5:25 AM
AFAIK, A GOP is defined by an I frame; a typical GOP would contain one I frame followed by a number of B and P frames.

I find the best use for increasing I frames when I'm rendering my video for DVD menus. When I'm using DVD Lab Pro I have a lot more control over where I place my cell breaks, since these need to happen on an I Frame - you can create some pretty killer seamless menu stuff.

--Scott