Practical DV FilmMaking

I frames are 'index' frames and contain the entire information in a frame. The codec uses these as a reference to compare with the next few frames.

Even here the codec performs different levels of compression on different parts of the frame, so that the centre of a frame is compressed less than the outer, permitting about 15 per cent reduction in the size. Next, the B frames - 'bidirectional frames' - are noted. An analysis works out the differences between a B frame and the next or the one before. It discards any data that is repeated, keeping a code of where it was. It is like compressing a numerical sequence that runs 22222333366666 by expressing it as 253465 (the digit '2' is followed by a digit representing the number of times '2' appears, and so on). In this way, a 14-digit sequence has been reduced by half.

Finally, P frames are labelled as 'predicted' frames and reduce data by predicting where certain objects are going to move across a screen. Once there is significant movement or change - after a half second or so, perhaps - the whole process starts again with a new sequence being labelled I frames at the start, then B frames and so on. Thankfully you don't really need to know how this all works, but it does prove useful later when we look at DIY encoding.

The good codec guide

H.263 - use for video conferencing

This codec is a good choice for the business community. It's ideal for a video conference where you do not need high-quality pictures and where audio is going to take priority. It is best with low move­ment films, so a more or less stationary talking head is going to be just perfect. The data rate for this would be small, but much quality is lost. Perhaps the only use of this codec is if you wanted to send a quick version of a film for viewing by a co-worker; you may not worry too much about the way it looks as you just want quick feedback.

Cinepak - use for CD-ROMs

This is a good, well-established system which works best with small image sizes. This codec is outper­formed by many others, even though its small picture size, at 120 X 90 pixels, is now getting bigger as computers improve. It is better used on low-end machines but is not usually the first option for web film, being more suited to CD-ROMs. A big advantage with this codec is that it allows the sender to cus­tomize settings throughout a film, so you can apply heavier compression to places where there