Intra VS Inter frame3 min read

Reading Time: 3 minutes

Since the introduction of the AVC-Intra codec, there has been questions on the difference between intra-frame compression and inter-frame compression, and how they compare to each other, so here’s a quick look at how they measure up.

There is no ‘better’ one of the two — it only depends on what aspect you are looking at. Inter-frame compression techniques, such as the popular MPEG-2, offer up small file sizes at ok qualities, whereas intra-frame compression schemes such as the new AVC-Intra or the JPEG 2000 codec offer very high quality footages at equally high file sizes.

To clearly understand how this is so, we have to look at how the compression works.

Intra
Intra-frame coding is a combination of lossless and lossy compression techniques performed only on the current frame; i.e. on each individual frame. To paint a clearer picture — each frame is compressed on its own (similar to when you save a jpeg file at lower quality in Photoshop), and has zero interactions with adjacent frames.

Intra compression is often known as the ‘I-only compression’ as every frame is an intra frame. That is to say, in a sequence of 6-frames, the sequence would be: I,I,I,I,I,I.

This process allows the footage to stand up well to footages with lots of motion, and also in post, where editing is easier (due to the independence of each frame). It also offers ‘native’ frame rate recording, whereby the image is processed frame by frame, then recorded one frame at a time.

There are several instances where inter-frame will show significant image deterioration when compared to intra, such as: News gathering scenes with multiple flashes; fast-paced sports coverage; confetti and rapidly changing lights (such as in events or MTVs).

Inter
The ‘inter’ in inter-frame compression refers to term ‘interframe prediction’; whereby individual frames are grouped and only the difference encoded. For e.g. in a scene where a car drives by a stationary background, only the moving car will be encoded. This would mean that each group of picture (GOP) will begin with one reference frame, and is followed by predicted frames, which utilises either motion compensation or image data processing.

That reference frame, of course, is what was mentioned earlier: an I-frame. Besides the I-frame, two other types of pictures make up the inter-frame compression scheme: the P(redicted)-frame and the B(i-directional)-frame.

A P-frame looks only at the previous I-frame, and from there generates the next picture. A B-frame looks at what precedes and follows it (be it a P or I-frame), and generates a picture from these two reference points. Following this format, a GOP lasting 12-frames can look like this: I,B,B,P,B,B,P,B,B,P,B,B.

It is then obvious that from this arrangement that the footage is difficult to edit, for the B and P-frames would not exist without the I-frame, and the I-frame may not always exist at the point which you want to splice. This predicament then drags your post machine into play — it will require much higher processing power to decode the inter-dependency of the frames. For example, to edit the inter-frame AVCHD footage on Apple’s Final Cut Pro 6, you will have to first transcode it into an Apple ProRes 422 Quicktime file before you can start editing it.

While intra-frame is often recommended for a quality broadcast-standard finish, many have learnt to work their way around it. Some news-gathering agencies rely on the small file sizes of inter-frame compressed footages for a quick upload and broadcast, and have structured their recordings so that the camera pans at a slower rate, and the background behind the anchor often times has less movement.