Activity for Stefan Fendt
-
Stefan Fendt committed [r613]
this file was rather useless to have on CVS ... (I really should take care of
-
initial CVS-upload of the filter (hope that works)
-
initial CVS-upload of the filter (hope that works)
-
this file was rather useless to have on CVS ... (I really should take care of
-
many bugfixes... friday-13th-effect
-
files in my working dir to be ignored by CVS
-
new changes (hope it works now) and ... new bugs...
-
mjpeg-tools use automake/autoconf , so yuvdenoise does...
-
*** empty log message ***
-
removed in favor of autoconf
-
the flickering-fixer was at the wrong position. It should have been just before
-
testing just other parameters (and slightly modified bad-good block matcher)
-
see ChangeLog-Comment
-
now 16x16 search but 8x8 blocks measured and exchanged
-
blocksize changed again ;-) now 4x4
-
forgot to restore the search radius ...
-
This hopefully speeds things up without affecting the image quality...
-
### WARNING ###
-
*** empty log message ***
-
Still unusable ... sorry... Small changes towards mjpeg_logging and new Y4M
-
Started nearly complete redesign (the only working part at
-
improved deinterlacer ...
-
Should run fine now...
-
Now MMX/SSE ... (psadbw is cool ... :-)
-
large modifications... Should work now like it is expected to...
-
MMX ;-) just say MMX ...
-
late night commit ... *too* many bugs left, but ... well, it works again ...
-
*** empty log message ***
-
*** empty log message ***
-
+++ Added search-radius-modulation !
-
Switched to cpu_accel to determine capabilities of CPU...
-
more options ...
-
*** empty log message ***
-
* better MMX/SSE support
-
loop unrolling ...
-
Should work now... I used EAX instead of EDX after cpuid,1 ... *sigh*
-
removed usage of mmx.h and switched over to real inline assembly...
-
fixes -lm ...
-
added -I .. deinterlace only.
-
half-pel search again ...
-
it was very late yesterday ... :-(
-
starting documentation for yuvdenoise...
-
another bugexchange...
-
bugfixes...
-
MMX(+e) halfpel-search
-
fixed -h bug ... Thanks Steven!
-
-
Well, this time some more changes:
-
updated ...
-
minor changes...
-
minor finetuning
-
finetuning...
-
removed unnesc. alignment (cygwin)
-
2nd try to bugfix segfault
-
Bugs:
-
next try...
-
deactivated experimental code...
-
Hopefully all oddities fixed ?!
-
get rid of this ...
-
trying to really get rid of test.m1v...
-
Ok, let's throw the rewritten denoiser into the cvs ...
-
Starting to get the new version into the CVS...
-
hopefully really kills it ...
-
minor changes --> removed warnings & better image quality...
-
fixed commandline-processing...
-
thresholding was a little to aggressive...
-
wrong default value ...
-
deleting gmon.out
-
initial commit... lots to do...
-
will remove this one soon...
-
initial commit
-
-
various changes... works quite well, now...
-
code clean-up...
-
near rewrite
-
better field-interpolation-code
-
there must be a bug in
-
Does work quite nice, now...
-
minor improvements
-
...
-
Hi,
-
OK, this version should be usable :)
-
better aa_interpolation...
-
- image variance fixed
-
added missing "-lcpuaccel" ...
-
hmm... disabled mmx... there's something weird with the libs.
-
*** empty log message ***
-
cleanup/speedup
-
code cleanup/speedup
-
code cleanup/speedup
-
modified blocksize, for a better result...
-
do not use this :)
-
The everything works, now release :)
-
bugfix :-}
-
this one avoids trouble with repeating patterns...
-
further speedups and better statistics for prediction of MVs
-
implemented a fast predictional zonal search.
-
trying to implement a more clever searching strategy...
-
code cleanup, part I (don't use this!)
-
bugfix: motion estimation works again... (turn's off some speedups)
1 >
