Image Processing Reference
as an MPEG-2 file and some pixel values have migrated from one field to the other during
the MPEG-2 coding process. This is a good example of why you should avoid re-encoding
footage that has already been compressed with another codec.
Note also that although we have eliminated most of the double exposure, we have
in fact lost a field of motion and the third frame closely resembles the fourth.
This technique eliminates interlaced motion artifacts, but for film-based content, it is
going to result in jerky motion and wastefully duplicated frames.
So this is not going to work with film-based content, although for normal video-shot
footage being transferred to the web it might be fine.
A more sophisticated and better method of de-interlacing uses an adaptive technique.
Adaptive processing treats only the moving parts of the picture. The stationary parts
are left alone. For locked-down shots this might be OK, but it's not very good on pans
unless the adaptive processor is sophisticated enough to detect the motion.
Technology has not yet been invented to simulate the effect of the human eyeball
in translating an interlaced image into a clean, progressively scanned format; however,
it is an area that computer science is working on all the time. The performance of video-
processing tools will no doubt continue to improve, and the resulting output will only
The major benefit of adaptive de-interlacing is that it effectively doubles the resolu-
tion when compared with the basic field-dropping approach.
There are some downsides to this technique, however, as it introduces errors with
some content. For example, anything that is gently and smoothly panning vertically is not
handled sympathetically by this approach.
Notably, this adaptive technique makes a mess of on-screen credits. The recommen-
dation by expert compressionists would be to retype the screen credits as stationary
Screen credits also get clobbered by the Gibbs effect (also known as mosquito noise).
This is a cross-color artifact that shimmers around areas of text.
It might be a good idea to apply a combination of techniques to separate parts of the
program and then join them up again.
This helps to explain why the preprocessing stage of a project accounts for such
a large proportion of the time you spend working on it.