Difference between revisions of "CELT TODO"

From XiphWiki
Jump to: navigation, search
Line 1: Line 1:
 
== Open issues ==
 
== Open issues ==
  
* Code
+
* Enable new layout with more HF bands?
** Laplacian energy encoding
+
* Prevent errors at very low bit-rates
** Variable spreading decision in the encoder
+
* Reduce waste when splitting at high bit-rate?
 +
* Remove 2-phase stereo?
 +
* fix start/end energy prediction
 +
* Revert N=1 allocation special case?
 +
* Revisit folding
 +
** Can we remove the need to fold on split?
 +
** Should we do L/R splitting?
 +
** Should we fold the side (probably not)?
 +
** Band-aligned folding?
  
 
* Tuning
 
* Tuning
Line 16: Line 24:
 
*** energy prediction parameters (means, decay)
 
*** energy prediction parameters (means, decay)
 
** Spreading/folding parameter
 
** Spreading/folding parameter
 
 
 
== Other decisions to make ==
 
* What should we do about remaining bits after the last PVQ?
 
** more fine energy?
 
** encode bands backward?
 
** allow "fractional pulses"
 
* Minimise bits left after PVQ
 
** Reverse band ordering?
 
** Fractional pulses?
 
* Dynamic allocation
 
** Include a way to deviate from the normal bit allocation?
 

Revision as of 11:01, 30 December 2010

Open issues

  • Enable new layout with more HF bands?
  • Prevent errors at very low bit-rates
  • Reduce waste when splitting at high bit-rate?
  • Remove 2-phase stereo?
  • fix start/end energy prediction
  • Revert N=1 allocation special case?
  • Revisit folding
    • Can we remove the need to fold on split?
    • Should we do L/R splitting?
    • Should we fold the side (probably not)?
    • Band-aligned folding?
  • Tuning
    • Band layout (spacing)
    • TF resolution options
    • Bit allocation
      • Main table
      • Stereo
      • qtheta offset (all 3 modes)
      • delta allocation (all 3)
      • fine offset
      • energy prediction parameters (means, decay)
    • Spreading/folding parameter