|
 |
In article <3ced21bd$1@news.povray.org> , "=Bob=" <bob### [at] threestrands com>
wrote:
> Anyone tried this?
>
> http://math.berkeley.edu/~benrg/huffyuv.html
On his website the author claims that using Motion JPEG would be inadequate to
produce MPEG because it would introduce artifacts. However, the opposite is
true. Conversion from Motion JPEG to MPEG is absolutely without quality loss
(if done properly - of course one can mess up anything if one does it
incorrectly) and it also much faster *.
Given that little knowledge of the author about the topic I seriously doubt he
is credible. Image compression is not physics ;-) His codec is only faster
because it does far less work which then the MPEG encoder will have to do
later anyway. Even worse, who knows what unexpected bugs or side-effects his
code has!
Deciding between some proven method like Motion JPEG and some method someone
cooked up obviously without even understanding much about the topic, well...
Thorsten
* The reason is that Motion JPEG as well as MPEG (1/2) share exactly the same
data structure and encoding/decoding algorithm for the DCT coefficients.
Consequently you don't need to perform the time consuming DCT/IDCT twice but
can simply transcode the image data. All good programs already support
transcoding of Motion JPEG to MPEG. The improper way is to first run a JPEG
decoder and then an MPEG encoder of its output. There are plenty of good
explanations about the exact process on the web to read for the boring
details.
Post a reply to this message
|
|
