From: Klaus Robert Suetterlin <robert@mp...> - 2006-07-13 09:16:28
this is a new idea I work on --- i.e. problem I'm unable to solve myself
:) ---, maybe someone knows if this is at all possible, or where I might
get information about it:
Our lab equipment produce two streams of mjpeg MOVs, each is PAL
standard, i.e. 50Hz interleaved, video. The source is a Sony
Progressive scan PAL camera, with two video outs. Each channel is
recorded as interleaved mjpeg.
The point is, that each video channel produces PAL standard conformant
video signals. But if You combine the synchronous half-frames of both
channels, You get a 50Hz non-interleaved (progressive scan) image.
Problem is: Currently we export both movies to some standard image
format, e.g. PNG. So we got two image sequences s0, s1, with images
si(j), with interleaved fields si(j,f). These sequences are
deinterleaved and then mixed into one image sequence like this:
s2(0,0) = s0(0,0)
s2(0,1) = s1(0,0)
s2(1,0) = s0(0,1)
s2(1,1) = s1(0,1)
s2(2,0) = s0(1,0)
s2(2,1) = s1(1,0)
s2(3,0) = s0(1,1)
s2(3,1) = s1(1,1)
s2(j,f) = sf(j/2,j%2)
Of course this is a lot of decoding images (jpegs of each field)
interleaving (into one image) encoding (into e.g. PNG) deinterleaving,
mixing and reinterleaving, and finally recoding into a movie.
In my opinion it should be possible to either rip the raw JPEG fields
from the original movies and then create a new artificial movie. This
would mean we do not have to de/encode video frames all the time.
But even better it should be possible to simply redifine a MOV movie
track with correctly mixed fields.
Unfortunately I do not know the internal format of MJEG, so I just do
not know if the second idea is at all possible.
Any pointers beyond the quicktime docs from apple welcome. Any more
help very welcome.