changes.encoder

MPEG-2 has 7 distinct parts as well. The first part is the Systems section which defines the contain

94/07/04 public release of version 1.1a
=======================================

94/07/04 sE

fixed wrong frame rate comment in .par files

mpeg2enc.c:
- fixed error message refering to wrong filename (argv[1] -> argv[2])

putseq.c:
- changed a .sxf to .syf

94/06/30 public release of version 1.1
======================================

94/06/30 sE

added missing _ANSI_ARGS_ in motion.c, predict.c, stats.c

94/06/28 sE

idct.c:
- removed short-circuit for even-coefficient-only case

94/06/19 sE

added non-linear mquant (cfogg)

conform.c:
- allow profiles >= 8, in this case no conformance checks are
  made, needed to create 4:4:4 sequences

stats.c:
- added output of # and percentage of coded blocks
- deactivated output of cbp, dct_type and motion vectors


94/06/17 sE  cummulative changes since version 1.0

supports 4:2:2 and 4:4:4 chroma_format
supports field pictures
supports dual prime motion vectors
moved conformance checks to conform.c
tighter conformance checks

fdct.c:
- modified rounding expression, this might reduce floating point
  implementation dependency, which caused verify to report differences

idct.c:
- see CHANGES.decoder

motion.c:
- moved macro block loop out of estimation routines
- variance is now always computed (even for I pictures)
- if no B pictures: estimate dual prime vectors
- added field picture motion estimation routines
- frame estimates returns all field vector estimations (needed for dual prime)
- changed fullsearch search pattern from line scan order to outward spiral
  (slightly faster when used with short-circuit dist1)
- dist1 speed enhancements: unrolled inner loop, stop accumulation if
  partial sum exceeds a given distance

predict.c:
- several modifications for chroma_format, field pictures and dual prime

putpic.c:
- moved mquant 'moderation' to ratectl.c
- several modifications for chroma_format, field pictures and dual prime

putseq.c:
- report correct numbers of I,P,B pictures to rate control (even for
  first / last GOP)

putvlc.c:
- fixed wrong DCT coefficient limit checks for MPEG-1
- added putdmv() for dual prime
- added support for cbp = 0

quantize.c:
- added correct saturation of coefficients

ratectl.c:
- changed some variables from double to int
- calculate spatial activities in advance
- improved VBV delay calculations and checks

stats:
- added output of motion vectors and coded block pattern

94/05/17 public release of version 1.0
======================================

94/05/15 cfogg

- renamed zz to zig_zag_scan, zz2 to alternate_scan,
  quanttab to default_intra_quantizer_matrix,
  qscaletab to non_linear_mquant_table
- additional comments


94/05/15 sE

style changes and some additional comments in all files

fdctref.c:
 - same PI guessing as in idctref.c


94/05/12 sE

name changed to mpeg2encode

readpic.c:
 - conv444to422 and conv422to420 now convert one component per call, not both
   at the same time
 - changed subsampling filters to be based on those proposed by Cheung
 - added filter for MPEG-1 4:4:4 to 4:2:2 subsampling


94/05/11 sE

mpg2enc.c:
 - changed definition of sample rate from width*height*picture_rate
   to horizontal_size*vertical_size*picture_rate
 - added display of version / copyright when called without parameters

global.h:
 - added version[] and author[] strings (similar to the decoder)


94/04/?? sE

first preliminary alpha release