Motion Codecs


by Abraham Macías Paredes


SourceForge.net Logo

Under the name of "Motion Codecs" I have tried to make lossless windows codecs that compress motion (not only every frame separately).
The logic is very easy. One frame normaly differs only a little from the previous so the actual frame minus the previous frame may result many zeroes. And the zeroes can be compressed more than other number.

The bad size is that this is not allways true, and this frames can't be compressed very much. So only fixed images composed videos or animation movies ("cartoons") can be compressed better with those codecs.

These codecs have been done using the huffyuv codec like a skeleton.

Predictor codec
This codec uses the "predictor" algorithm as decribed in the RFC 1978.
Is a fast lossless codec but with very little compression.

Mhuffyuv codec
This codec is a modification of the original huffyuv codec to use "motion".

Direccional codec
This codec uses the "direccional" algorithm described here.


Comparative Chart:

The test video has been created from http://www.biteycastle.com/biteyWindow.htm (a flash animation called bitey of brackenwood) and a shareware version of ADShareit SWF to Video Converter. Only a few seconds of the start of the movie were used and with no sound.

CODEC
 FILE SIZE
 RAR COMPRESSED
CorePNG
 30 MB
 29.3 MB
CamStudio
 31.8 MB
 31.4 MB
AVIzlib
 36.5 MB
 34.6 MB
Direccional
 47.5 MB
 26.9 MB
MindVid
 59.4 MB
 27 MB
VBLE
 71.2 MB
 28.7 MB
Predictor
 86.6 MB
 22.8 MB
AVImszh
 91.5 MB
 32.5 MB
LOCO Codec
 101 MB
 36.8 MB
mhuffyuv
 120 MB
 34.3 MB
PICVIDEO Lossless JPEG
 125 MB
 25.6 MB
huffyuv
 148 MB
 36.9 MB


As you can see the best compression ratio was with codecs using the zlib. My next step will be create a new format applying the improvements of the direccional codec to the predictor codec in order to archieve a fast codec with a good compression ratio. And then I will try to use the Motion Compensation of the mpeg video with the direccional codec or the predictor with improvements (the best of the two).


NOTES: