MJPEG Tools Svn

remove lingering cruft (HAVE_STDINT_H) from pgmtoy4m's prior existence outside mjpegtools

clean up, remove unused cruft from mjpeg_types.h and win32defs.h, HAVE_STDBOOL_H goes away since none of the C code uses it (and C++ has it builtin). mjpeg_types.h is a shadow of its former self ;)

using 0 and 1 obviates the need for stdbool/bool - the HAVE_STDBOOL_H can go away now.

Missing prototype for posix_memalign caused a compile error on system which lacked that function.

minor improvements

Add a bit more explanation about LPCM

Fix mis-sizing of b-group's based on min not max GOP size

Fixed incorrect B dropping for closed GOPs.

Hi,

* taken anti-aliassed-scaler out of work, again
as it seem's like the sinc-scaler does visually
fit better to the motion-search (!?!..?! Huh?)

* better vertical motion-vectors. It seem's like
I have found a way to come arround the positioning
error of +/-1 in most (not all) cases.

* two new modes (switched *after* MC) to do it even
better(?): Progressive-detection and Telecine-detection
(and removal)

* minor bugfixes(-exchanges) and speedups

have fun...

PS: is anyone using it? If so, could you please run it
against your favourite interlaced stream and tell
me about the result?

when creating an output file don't set the 'read' flag - libquicktime does not
do the right thing if the file does not exist