>
> Denis Krasnov (dkrasnov_at_nyc.rr.com) wrote:
>
>> The following is a part of the same movie which I cropped, turned and
>> stabilised
>> Now ,does it jiggle or not ? Remember we're talking about position of what
>> is on the screen relative to the rest of the unit.
>> Watch
>> www.mentalemetic.com/other/newton/00.mov
>
> Could you get this just a tiny bit smaller?! (It's about 37.5 MB now.)
>
> My "enhanced" version is just 130 KB -- I merely copied a typical
> sequence of 10 frames and cropped the movie to the relevant area. Scaled
> to twice the size and viewed in a loop, it becomes quite obvious that
> yes, the text retains its position while everything else moves. Here it is:
>
> http://www.michael-hussmann.de/iwalk.html
Right , now I see the origin of your confusion
Watch this
http: //www.mentalemetic.com/other/newton/say_croppedx400.mov
This is a piece of movie you posted magnified 400% and cropped
Now watch this
http: //www.mentalemetic.com/other/newton/say_croppedx400_good.mov
it's absolutely the same segment of the original movie (0.mov) after I
decoded it using proper software, and then magnified and cropped
Both this movies are now saved without lossy compression ( as well as my
00.mov was , what was the reason for its size)
This is essential.
The original movie
http: //www.mentalemetic.com/other/newton/0.mov
was saved in cinepac compression, this sort of mpeg like compression is not
pixel and frame accurate. What it does it puts key frames ( some times as
far as 2-3 seconds apart) then it encodes those frames only, using some jpeg
like algorithm, which requires the image to be sliced in square blocks (
which could be as big as 32 pixels). These blocks are then encoded via
Fourier Transformation. Frames between key frames are encoded and decoded
applying motion vectors, which means finding blocks that moves as whole from
frame to frame without changing its infrastructure and encoding the
information about its relative velocity instead of encoding pixels.
If encoder or decoder discards some of this information you get precisely
the sort of fluctuations you pointed out ( one of motion vectors hasn't been
applied and a part of the image is seen in absolute rest for a moment) .
This is an artefact, and though it has probably originated during encoding
end only reinforced then you decoded it, it appearance can be minimized by
using more accurate decoding techniques ( this is what I've done).
Now, to see an irrefutable proof that it was an artefact for yourself look
at the letters "He..", which are being written at the same time. They appear
to be at a lock with the rest of the screen, to put it in other words only
the word "Say" jiggles , not all of the text. The contrary would be the case
if the text was overlaid, to put it differently, if the text was overlaid it
would appear to jiggle as one - the "He" would jiggle too.
Cheers, Denis
P.S. This is what I meant by "As far as noise due to video compression
allows to see" in my original note.
>
>
> Michael J. Hussmann
>
> E-mail: michael_at_michael-hussmann.de
> WWW: http://michael-hussmann.de
>
-- This is the Newtontalk mailinglist - http://www.newtontalk.net To unsubscribe or manage: visit the above link or mailto:newtontalk-request_at_newtontalk.net?Subject=unsubscribe
This archive was generated by hypermail 2.1.2 : Fri Feb 01 2002 - 16:02:15 EST