Timestamp-independent motion vector prediction for predictive (P) and bidirectionally predictive (B) pictures

A method for use in encoding video data, the method comprising:

• within a sequence of video pictures, selecting a current video picture to be encoded;
• dividing the current video picture into portions and selecting a current portion to be encoded;
• establishing at least a first reference picture for said current portion;
• and
• selectively assigning at least one motion vector predictor (MVP) to said current portion, said MVP including data associated with at least said first reference picture and with at least one other encoded portion of said current video picture, and wherein said MVP is not based on a temporal interpolation of motion vectors used for encoding said first reference picture, and wherein selectively assigning said MVP to said current portion further includes:
• encoding said current portion using a Copy Mode scheme based on a spatial prediction technique to produce a Copy Mode coded current portion;
• encoding said current portion using a Direct Mode scheme based on a temporal prediction technique to produce a Direct Mode coded current portion;
• and
• selecting between said Copy Mode coded current portion and said Direct Mode coded current portion, wherein selecting between said Copy Mode coded current portion and said Direct Mode coded current portion is accomplished using a Rate Distortion Optimization (RDO) technique, wherein said RDO technique uses a Lagrangian parameter λ based on a quantizer (QP) associated with said current portion, and wherein said RDO technique employs an adaptive weighting function, wherein said adaptive weighting function includes:
• f
• ⁡
• (
• QP
• )
• =
• max
• ⁡
• (
• 2
• ,
• min
• ⁡
• (
• 4
• ,
• QP
• 6
• )
• )
• .