IBP vs IBBP vs IBBPBBP?? What format to solve this problem?

[Dreath]

Ok, can someone tell me the trade offs between those three GOP setups for MPEG encoding?

I'm trying to get good compliant MPEG2 files of TV shows at 352x240 VBR 3500max to make simple DVD's of them.

I recently recorded Battlestar Galactica from the Sci-Fi channel and when I used Wombles MPEG2VCR program to scan for errors, it came up with over 10,000 GOP time code errors!!!! That is about 2-3 errors a second!!!

I'm using an ATI TV Wonder PCI with a AMD Duron 1.1GHz and 256megs SDRam under WinXP pro. I'm using PowerVCR II for recording the shows which so far hasn't given me any issues.

However, because of all these errors could it be related to CPU overload while trying to record the movie to MPEG2? If so, would changing the GOP frames set from IBBPBBPBBPBBP (i think) to something like IBBP or IBPBP would be better?

Please help as I'm doing a crash course in MPEG file structure to understand these problems.

[lordsmurf]

I'm gonna go from memory, so hopefully I won't screw this up...

The xPEG system (MPEG, JPEG, MJPEG) are all based off the idea of "what has changed since the last one" with motion images, as well as "how many colors can I clump together in the file" in still images or in each frame of the motion image

So instead of 111111111122222222223333333333 the data would be shorthanded as 1x10,2x10,3x10 ... because there are 10 of each number .. that's still and single frames from motion. Now with motion, you also have 1122334455 and the next one is 01122334455, the only thing different is 0 ... so 0 is what is recorded as the change, others left alone and refer back to previous data ... granted, it's all more complex than this, but easy way to explain

======

For the specific question:

I is the primary
then P is based off I
then B is based off P

So the more B and P you have, then more compression is introduced. Pressed DVDs most often use IBBPBBPBB as the range ... I try to closely match that unless good reason exists to further compress or uncompress ... and in the case of my ATI AIW cards using ATI MMC, I use a smaller radius to increase quality slightly

Before wasting time on scanner results, ask yourself... does anything even look wrong ... and let's not forget 352x240 MPEG2 is not a very desirable format ... look at using 352x480 instead

[Dreath]

Thanks for the info. I'm just curious if there would be compatability problems if I tried to use a IBP format compared to a IBBP format. I know the IBP will be bigger in size but the quality will be better.

As for the scanner, it might be that it doesn't like the 320x240 size. Who knows. The reason I chose it was that I'm planning to play it on my TV and I don't want the video to look like it's been stretched too tall.

Which is what confuses me. If most DVD/VCD/SVD formats use video ratios of 4:3 or 3:2, why would you suggest 320x480 which is a 2:3 ratio???? I'm confused...

[bugster]

Which is what confuses me. If most DVD/VCD/SVD formats use video ratios of 4:3 or 3:2, why would you suggest 320x480 which is a 2:3 ratio???? I'm confused...

TV pixels are not square. You can't take the X and Y of digital resolution and use it to calculate the aspect ratio.

DVD is usually 720 * 480 for NTSC and 720 * 576 for PAL (Full D1 resolution). Either of these can be displayed as 4:3 or 16:9.

[Dreath]

So Bugster, what you are saying is that I can record the video at 352x480 resolution and play it back on my TV and it will look fine?

However, I bet if I play it back on my computer monitor, it will appear as though the video was stretch verticle. Am I correct in this observation?

[lordsmurf]

So Bugster, what you are saying is that I can record the video at 352x480 resolution and play it back on my TV and it will look fine?

However, I bet if I play it back on my computer monitor, it will appear as though the video was stretch verticle. Am I correct in this observation?

Looks fine...

Yes for tv.
Yes or no for computer, depending on playing software. Software should respect aspect ratio. Windows MP will not usually. PowerDVD and WinDVD will

[adam]

I frames store all of the redundant information. Picture a simple movement occuring over 1 second. Lets say a man is standing in front of a building and in that 1 second his hand moves 1 inch. The I frame would just be a still picture of him standing with his hand in the starting position.

P and B frames fill in the gaps between each I frame. So in this example, the P and B frames would ONLY contain the information for the hand movement, because the man and everything behind the man are the same for each frame, and already stored in the I frame. These P and B frames reference the I frame, taking out all that info and then essentially just tack on their little bit of info to create the new updated image. When you go through each image in sequence, you get movement and its much more efficient to do it like this than to store the entire picture over again each time something little changes.

Now all you asking in your question is whether it is better to have alternating B and P pictures, or twice as many B pictures as P ones, and the answer is that it depends.

As you can see from my above description, mpeg encoding is all about finding redundancy and removing it. P frames are called predicted, because they look BACK to previous I or P frames and only store those macroblocks which differ, ie: the hand movement.

B pictures are called bidirectional because they look both forward AND backward to I and P frames. So efficiency of THAT particular frame is better and thus higher quality, but remember B and P pictures only look to I and P pictures. Nothing ever looks to a B picture for information. Once you encode a B picture that's it, nothing will ever be able to reference it for information. Its a one shot deal, so while that B picture will be higher quality then if it were a P picture, future frames will suffer. A good analogy is money. You can spend some now making that day better for you, or you can save it and let it gather interest so that future days will be better.

So here is the short answer...at typical to low bitrates increasing the number of B pictures can decrease SNR (signal to noise ratio) which means better quality. The downside to this is increased bandwidth, but as long as you stay within the limits of your respective standard (DVD, SVCD etc..) then your hardware/software shouldn't have any problems.

But as your bitrate increases, B pictures become less important and are essentially just wasted. So IBPBP has the potential to produce higher quality, but you need significantly more bitrate to realize it.

So with all that said, I agree with Lordsmurf's conclusion. Stick to IBBPBBP.