Yes, it was a bit of a long post..... and I wonder if I could perhaps put the case 'for' occasionally shifting up from 1080i60 (30 fps) to 1080p60:
1080i is really 1080i60, ie 60 half-frames (fields) per second and 30 frames per second. On TV's the interlacing is smoothed really well, so both the still parts of the image are very sharp (as sharp as 60p) and the motion is very smooth smooth, because there are 60 changes per second where things are moving. TV's use custom adaptive deinterlacing algorithms to achieve great results. Computers, generally, don't.
What's wrong with converting 1080i60 to 1080p30? Well, the two half-frames have to be joined together somehow to make one progressive frame. Speaking generally, there are three kinds of solutions:
1. throw out one of the half-frames completely, and stretch each pixel vertically of the remaining frame so that it occupies 2 vertical pixels. After doing this, you have 30p, and each individual frame will actually look surprisingly good (up close diagonal lines will be jagged, etc). Motion will be more jerky than 60i on TV because half the movement information is lost, and the vertical resolution is halved even in the still / non-moving parts.
2. mix both frames together, by merging each pair of vertically adjacent pixels (one from each half-frame) together. This blurs any object that moves, and again reduces vertical resolution by half. A black object moving over white will end up with grey blurs on each edge. This doesn't look as jerky as option 1, but doesn't look great, and still frames look really bad.
3. use a 'adaptive' deinterlacing algorithm to keep as much information as possible from both fields. Where there is no movement at all from half-frame to half-frame, all pixel information is retained, and the output is a progressive image that contains all the information from both fields. In regions of movement, the algorithm attempts to make the best possible image from both fields, typically using the data from one field only, so that where there is movement, the vertical resolution is halved. Movement will still be jerky, even though individual still frames will be the sharpest of any of the three options.
In all three of the above options, movement tends to be jerky or blurry, because moving objects are not updated every 60th of a second, instead only every 30th of a second.
So that's why sometimes going from 1080i60 (30 frames per second) to 1080p60 can be useful. There are two ways to go from 1080i60 to 1080p60.
1. For every field of the 1080i60 original, spread each pixel vertically to be twice as high, making one frame from each field. So you end up with 60 frames per second from 60 fields per second. Each new frame has all the horizontal data but only half the vertical data. The end result is video with 60 different frames every second, motion is really smooth (as smooth as 'real' 60p), but the vertical resolution is half that from a 'real' 60p sensor, even if there is no movement.
- there is an adaptive way of doing this as well, and the algorithms can be very complex indeed. Basically this is what TV's do. In the simplest solution, where there is no inter-frame movement, the data from both fields are written on alternate lines of a progressive frame to make a frame that is as sharp in all respects as a 60p original. Where parts of the frame are subject to movement, a smart algorithm decides what would look best, either doing some kind of blurring or using vertical pixel stretching that retains as much detail as possible in the non-moving parts of the image.
Both of these solutions give full smooth 60fps movement while retaining a lot of the detail, the latter giving frames without movement that are as sharp as if the same scene was shot at 60p, and because movement sometimes looks better if it is a little blurred, these results can be very good indeed (depending on the algorithm).
So sometimes it does make sense to go up from 60i to 60p, especially where smooth movement is important, and if the camera is relatively still for quite a lot of the time. Depending on how it's done, quite a lot of detail can be retained. I've had really good results with JES Deinterlacer.
If the underlying quality of the video isn't great, going from 1080i60 to 540p60 (960x540) is an option, especially for playback on iPads or iPhones. This keeps all the vertical and movement information from the original 1080i60 file but loses half the horizontal information. The end result is very smooth motion in a compact file that is only a little smaller than a 720p60 movie.
I like the idea of RevolverHD. If it works, it might be quicker and easier than using ClipWrap, and cheaper. I'll give it a try.
Chris.
PS - I actually reckon that for 'prosumers' FCPX is a huge step forward from both iMovie and FCE. I'm quite enjoying using it.
