How is H264 interlaced handled?

[LariosoDog]

Hi

Running 1920x1080 all the way - but wonder about what is usually available 24p or 50i.

H264 has predictive frame in between I-frames, but how is this done with this kind of encoding.

If you are to encode just every other line you will get a lot of change on every P-frame.

Or is every frame presented and encoded the same, and player just picks every other line for presentation?

Like if 25p original, that you encode that H264 25 fps and it is up to player to pick each frame at double speed and presenting top line frame first, and then next frame 2nd line down every other line.

I like the smoother look on 50i, since I run PAL - but like it is tiny bit more grainy.

So would be great to know if any good reason for that.

Thanks.

[Bill Sprague]

What camera are you shooting with and what are you setting it at?

[LariosoDog]

I get 25p i general, no interlaced from camera. I can run 50p on one camera but have not tried that, but will probably use to get smooth slo-mo at some point.

So I get normal H264 with predictive frames or the Canon All-I that is just H264 I-frames all the way as I understand it.

I render for AVCHD only, either on DVD or memory stick of some sort.

All is 1920x1080. Large majority currently is time-lapse, but will be normal video shooting later too at some point.

In some editors I can choose whatever at rendering, progressive or interlaced.

Since I can only select 23.97, 25 or 29.97 progressive in PRE I have tested in other editors to select interlaced 50 fps and like that better especially if doing pan with video. Anything else than camera do, 25 fps, always get some anomalies in movement.

Don't know if AVCHD put restrictions on bandwidth, not quite like bluray as I understand it.

I think you only get 24-30 fps progressive, or 50-60 fps interlaced.

I am experimenting a lot to learn what gives what result, both shooting and editing wise.

And in the end on Vimeo will test what they do to the matter - when I know how it works offline - and what give best results.

So now trying to figure out how H264 works if doing interlaced - will it try to encode odd and even lines separately as two different frames?

Or is it players job to double framerate if interlaced, and pick in two sets of lines/rows - and full frame is encoded once?

[Bill Sprague]

I'm sorry, but I don't understand your question or even if you are asking one.  

My understanding is that "interlaced" had a good purpose when we were using tube TVs and watching low resolution broadcast.  It might be better to ignore it from now on. 

My copy of Premier Elements has so many output settings I'll never understand them all. 

My personal pattern is to shoot everything "progressive".  It was 1920x1080p for a few years and is now 4K at 30p.  When a project is complete, I output using a custom option and take advantage of the "Match Source" button.  I put a copy on a flash drive for my TV and upload a copy to Vimeo.  Vimeo seems to have no trouble with what I send it.  ( Bill Sprague on Vimeo )

Frankly, unless you are using unusual cameras or footage from unusual sources you may be trying to make this too complicated. 

Good luck with your projects!

Bill

[LariosoDog]

Ok, thank you for tips and best wishes.

Yes, it's technical, more than most are interested in I suppose.

Cameras now are Canon EOS 80D and 100D and do progressive frames, no other choice.

Rather set and pleased with choice of cameras and lenses.

I am trying to figure out where to put more investments - better editors or authoring tools doing more quality rendering and have options enough to get the most from what you own - and which part of end quality is actually due to presentation like vimeo+player+monitors which I cannot control what viewers have.

This is one part of question - put more money into authoring tools to get enough options and quality?

Here is where it gets technical beyond most peoples interest, so I warned you.

H264 is a series of frames, where a group of frames(GOP) are first a full frame(I-frame) compressed, then maybe 15 or so predictive frames(P-frames) just containing differences from previous frame. Then it starts over with the same series with a new full frame. In detail this is MPEG-4, and H264 also look in next frame to come deciding how to compress(think it's called B-frames).

So doing progressive where all rows/lines in each frame is presented at once - nothing unclear.

Every decoded frame is a full image at that instant. Last I-frame plus the predictives you are on now.

Interlaced though will do half a frame at a time, odd lines separate in a frame, and even lines in next - can be set in some editors if top frame or bottom frame comes first.

And on same bandwidth - interlaced can then use double framerate - each frame is half size - in 1080 each interlaced frame would be 540 lines.

TV is actually broadcasted like this, first one half frame then next - for PAL 50 fps.

But how does the encoding process when interlaced?

Is video interlaced different than tv broadcasts?

Just imagine that doing predictive frame compression is very different deal compared to if each frame processed is the full image - or just the odd or even rows.

So wonder about if there is a technical reason for interlaced to look bit more grainy - not hugely but tiny bit?

This was the question - if anybody remotely can get what I try to ask?

Just a reference to a book title would be great.