| View previous topic :: View next topic |
| Author |
Message |
Brenton Chapin
Guest
|
 Posted: Fri Aug 08, 2003 11:53 pm Post subject: segue problem |
 |
|
This segueing problem has been known for some time. But it>s a fairly
minor problem with practical workarounds such as adjusting the
position of breaks to match the divisions in the encoding method, or
adding some overlap, or keeping all the data in one file and indexing
to positions within that file, or doing a very fast fade. I think the
Ogg Vorbis audio format addresses this problem. There>s an archived
discussion at http://www.xiph.org/archives/vorbis/200008/0047.html
Haven>t checked, but expect JPEG has the analogous problem with
images. Take an uncompressed image and cut it into 2 parts. Make the
cut at some number of pixels that is not a multiple of 8. Convert
each part to JPEG, then tile them back together. The join will be one
long visible artifact. |
|
| |
|
 Back to top |
Thomas Richter
Guest
|
| Posted: Sat Aug 09, 2003 1:33 am Post subject: Re: segue problem |
|
|
Hi,
[quote]Haven>t checked, but expect JPEG has the analogous problem with
images. Take an uncompressed image and cut it into 2 parts. Make the
cut at some number of pixels that is not a multiple of 8. Convert
each part to JPEG, then tile them back together. The join will be one
long visible artifact.
[/quote]
That>s a general problem for other image compression methods as well,
for example if you use tiles in JPEG2000, you>ll get artifacts along the
tile boundaries for high compression rates for the same reason. A "fix"
for this problem exists in JPEG2000 part 2, namely by making the tiles
overlapping and reconstructing to an average in the overlapped region.
As simple and obvious this may sound, the only reason for this not being
in the royality-free part-I is that IIRC, some company had a patent on
this "technology". /-: Wierd world...
So long,
Thomas |
|
| |
|
| Back to top |
Joe
Guest
|
| Posted: Thu Aug 28, 2003 10:55 am Post subject: Re: segue problem |
|
|
Thomas Richter <thor@math.tu-berlin.de> wrote in message
news:3F340906.8030801@math.tu-berlin.de...
[quote]Hi,
Haven>t checked, but expect JPEG has the analogous problem with
images. Take an uncompressed image and cut it into 2 parts. Make the
cut at some number of pixels that is not a multiple of 8. Convert
each part to JPEG, then tile them back together. The join will be one
long visible artifact.
That>s a general problem for other image compression methods as well,
for example if you use tiles in JPEG2000, you>ll get artifacts along the
tile boundaries for high compression rates for the same reason. A "fix"
for this problem exists in JPEG2000 part 2, namely by making the tiles
overlapping and reconstructing to an average in the overlapped region.
As simple and obvious this may sound, the only reason for this not being
in the royality-free part-I is that IIRC, some company had a patent on
this "technology". /-: Wierd world...
So long,
Thomas
[/quote]
There>s a patent attorney at this very moment trying to figure out how to
sue all the people reproducing (multiple meanings) the patented genes along
the lines of the ones for making insulin. :P |
|
| |
|
| Back to top |
|
