MJPEG Tools Svn

very ugly code but works...
the code is prepared to take any chroma-subsampling
the code is slow (mainly because of a slow MC)
it only works (currently) for bottom-field-first 4:2:0 PAL
the regained fields are of the best quality I ever have seen by any deinterlacerso far... (does anyone have a Trimension at hand to test against?)

Stefan

rename/relocate lavtools/lavtc.bash to scripts/lavtc.sh and update Makefile.am.
The script can always be removed from the new directory as well as the old ;)

Ding Dong the XYSCSS tag is (finally) dead. Smilutils has been updated to
not emit XYSCSS= and a patch has been submitted to ffmpeg to remove it from
libavformat's yuv4mpeg2 output.

Apparently some (win32) compilers are on the dumb side and don't know that
'unsigned int' and 'uint32_t' are the same thing. Thus calls to pvariance()
were having problems because the 3rd and 4th args are uint32_t but lum_mean
and lum_variance were 'unsigned int'. Changed lum_variance and lum_mean to
be uint32_t.

Remove misleading comment - the output is not 4:2:0 mpeg1 ;)

Complete the job of bumping the version to 1.9.0 (forgot the AC_INIT).
Bit of cleanup in yuvdenoise - use the y4m_chroma_ss_{x,y}_ratio calls instead
of hardcoding the subsampling factors. Use the y4m_chroma_description() call
to get the string version instead of hardcoding it.

raw pcm capability (-R). Courtesy of Burkhard Plaum

1. Cleanup - since ATTRIBUTE_ALIGNED_MAX was no where defined ATTR_ALIGN was
a NO-OP. Define ATTR_ALIGN to actually DO something and then inline it in
mmx.h (and remove attributes.h and all references to it).
2. Add the callback interface from Burkhard Plaum
3. Bump the version number to 1.9.0 (the divergence from 1.8.0-release has
begun ;)).

This commit was manufactured by cvs2svn to create tag 'RELEASE-1_8_0'.

Fix for 1.8.0