Advantages of using I-frame only?

[telemike]

I read that someone here is capturing MPEG2 with I-frame only. Will this give a better capture?

[Thorn]

When I had an extremely slow cpu and was trying to capture mpeg (also no HD space) I had to capture mpeg1 files in I-frame mode only. The CPU doesn't have to work as hard do do I frames so you don't drop frames as much. However, I-frames are less compressed so the picture quality is lower for a particular bitrate (so I cranked the bitrate up).

Now I have a faster CPU but I don't capture mpeg at all... I capture in huffyuv and then encode with TMPGenc and get a MUCH better result.

[BJ_M]

some pro applications (high end) capture in I frame only so the file is easy to edit ... but they are capture in very high bit rate ...

[SLK001]

However, I-frames are less compressed so the picture quality is lower for a particular bitrate

Thorn, this statement escapes me. What do you mean by this?

[patrickm]

i.e. a 1 Mbps IBP stream would look better than a 1 Mbps I frame only stream

I frames are basically JPEG impages. B frames and P frames contain only fragments of the picture that differ greatly from the previous I frame. for example, a typical IBP sequence might have 1 I, 4 P, and 9 B frames: IBBPBBPBBPBBPBB ... (repeats)

and takes up 150k (just for argument's sake). the I frame is prob 35k, the P's 20k each, and the B's fill in the rest by encoding only differences from the "high quality" 35k I frame.


those B frames are very small compared to I and P. if you used all I frames:
IIIIIIIIIIIIIII ...

then each picture would be a 10k I frame (full still picture). and each of the first stream would be similar in quality to a set of 35k JPEG's with some minor degradation, while the second would be a set of 10k JPEGs.

in general, I frame only streams are good for editing, and for similar picture quality will be about 2.5-3x larger than IBP streams.

(and in reality, for SVCD, the bitrate is about 11k/frame, so this rough example isn't all that inaccurate)

[Thorn]

There's probably a more in-depth explanation of mpeg somewhere around here, but here's the general idea:

Mpeg streams consist of I, B, and P frames. For the same quality (more or less) of image, a movie consisting of I, B and P frames will take considerably less space than one using only I frames. This is because the P and B frames can store images in a smaller space than an I frame could store the same image.

Now, on the other hand, if you are capturing at a specific bitrate (say, 3Mbps, or better 6Mbps) then your quality will be affected by your choice of IBP or I-frame only for capture. Since there are a limited number of bits available, additional picture information is discarded when there is no room left. IBP encoding (note that there are many more options than just on-off for I, B and P frames) leaves more room for the additional picture information -- hence a better picture.

Think of it like packing a suitcase: if you fold your clothes you can fit more of them in. If you just ball them up you can get them in faster (but they'll get wrinkled and you can't fit as many).

Ok, so my explanation is pseudo-technical and not fully accurate, but hopefully it helps a little.