changes_v60.txt

Mobile IP VCEG的信道模拟程序

Changes from version 5.9 to version 6.0

Encoder:
 - possible picture formats increased from (QCIF,CIF) to
   (Sub-QCIF,QCIF,CIF,4CIF,16CIF,CUSTOM).
   If CUSTOM is choosen, width and height must be devidable by 16
	- all picture size dependent buffers are dynamically allocated in 
          get_mem4global_buffers() and partly in init_img()
	- additional picture size parameters are added to config.dat 	

 - re-structured code according to VCEG-M52.doc 
   "Test Model Document Changes for Data Partitioning and NAL Support"

 - Context based adaptive binary arithmetic coding CABAC included
   switch between UVLC and CABAC in config.dat
   for further details see VCEG-L13.doc and VCEG-M59.doc
  
Decoder:
 - re-structuring has just started, see section "known problems"
    

Known problems / open issues:

  1 The current decoder doesn't work, as we have already started to change the decoder.
    Therefore the TML59 decoder binary is included as \bin\ldecod59.exe.
    This decoder is only tested for the following TML6 encoder mode combination:
    (UVLC && no data partitioning && no slices)

  2 The picture header has to be changed to enable the coder to signal the new picture formats
    -> only QCIF and CIF format is decodable, results for other formats only at
    the encoder side available.
  
  3 The encoder crashes for ((format > CIF) && (# of reference frames != 5) && (B-frames)) 
    this may be related to issue 8.
  
  4 B frame mode is not very stable, there might be other mode combinations including
    B-frames that don't work.

  5 In VCEG-M52.doc the following changes are suggested, but up to now haven't been
    incorporated in the software.
    "The implicitly given decoding routines in the encoder will be identified and modularized.
    By transferring these routines to the decoder software project a more consistent
    realization of the encoder and decoder is achieved.
    This will be important for future integration processes since it may reduce the risk of
    encoder/decoder mismatch and, at the same time, limit the burden of a duplicated implementation."

  6 The introduced modular software structure is implemented from the top level down to
    the macroblock level. It should be considered to have a modular structure for the block
    level as well.

  7 Some of the global buffers may not be necessary, e.g. different temp buffers
    used in different functions may be merged to a single buffer.
  
  8 It would be better to use just one buffer for all reference frames and not one
    additional for B-frame calculation (mref[no_multpred+1] instead mref[no_multpred]
    and mref_P. 

  9 Other color formats than 4:2:0 and color depths than 8 bit/pel and component should
    be supported. Related input parameters are included in config.dat, but currently
    have no effect on the encoding process.

Contact: [e-mail]
         [e-mail]
         [e-mail]