[XviD-devel] OBMC decode support patch for dev-api-4

suxen_drol suxen_drol at hotmail.com
Thu May 8 22:26:24 CEST 2003


On Wed, 7 May 2003 14:56:54 +0200 (CEST) Christoph Lampert <chl at math.uni-bonn.de> wrote:
> Does XVID support ARTS (= Advance Realtime Simple)? In how far? What is
> missing?

arts == simple profile + rrv + newpred.
we dont currenlty support newpred; i'll explain why:

arts is intended for situations where the encoder output is directly
coupled to a decoder (e.g. realtime streaming).  newpred is used when
the decoder detects an error in the bitstream. the decoder sends a
message to the encoder, informing the encoder of the last frame it
decoded without error.  on the next frame to be encoded, the encoder
uses that "last decoded frame" as the reference frame.

unfortunatly, the mpeg-4 spec doesnt define:
1) a limitation on the number of reference frames the encoder&decoder
must maintain. rather it seems, the decoder is permitted to buffer as
many frames as it wishes. if the encoder recieves a message indicating a
frame outside its buffer, the encoder ignores the message and the
decoder must wait from the next ivop/keyframe.

had mpeg4 provided a max_newpred_buffer field, i think the improvement
offered by newpred and rrv would have rival'd asps qpel, gmc and bframes.

2) decoder->encoder messaging format is not defined within mpeg4 visual
iso. however i believe its describe in the rtp handling specification.

cya
-- pete




More information about the XviD-devel mailing list