Deinterlacing
Also known as: interlace removal, de-interlace
Converting interlaced video, which stores alternating half frames, into whole progressive frames that modern screens can display properly.
Updated
What it means
Interlaced video records half the lines of each frame at a time, alternating between odd and even. It made sense for broadcast tubes and makes none for a modern display, which wants whole Progressive scan frames instead.
Shown without conversion, the two halves of each frame appear together and anything moving develops a comb of horizontal teeth along its edges. Deinterlacing is the step that reassembles them properly.
It matters for old material rather than new. Action cameras have always recorded whole frames, so this is a problem for archive footage, old camcorder tapes and the occasional broadcast clip.
How it works in practice
The best conversions interpolate rather than discard. Simply dropping one field halves the vertical detail, while a motion aware method along the lines of Optical flow reconstructs the missing lines and keeps what was recorded.
Field order has to be right or the result judders. Getting it backwards produces motion that steps forwards and backwards, which looks like a frame rate mismatch rather than a field order problem.
Doing it once, early, is the rule. Deinterlacing after any scaling or Transcoding works from already damaged input, and the combing gets baked into the very step that was meant to remove it. Convert before anything else touches the clip.
What to watch out for
Interlaced material at a nominal frame rate actually carries twice as many moments in time, so a conversion can either preserve the smooth motion or the frame count, and not both.
Mixing converted and native footage in one sequence shows up in motion rather than in detail, because the two have genuinely different temporal characteristics. Converted footage carries a smoothness that native progressive clips do not.
Automatic detection also guesses wrong on static shots, where there is no movement to reveal the fields, and the setting it picks then fails on the next shot with Motion blur in it.
How RawClip handles it
Modern action and phone footage is progressive, so this almost never arises. Where an interlaced clip does turn up in an upload, it is converted once before anything else reads it, rather than being analysed with combing still in the frames and scored down for damage it does not have.