Systems and methods for deriving a motion vector prediction in video coding
A method of performing motion vector prediction for coding video data is disclosed. A full precision motion vector mv is determined for generating a prediction for video block in a first picture. A rounded motion vector rmv having a less precision than the full precision motion vector mv is stored. A motion vector predictor candidate is generated for a video block in a second picture from the stored motion vector.
1. A method of performing motion vector prediction for coding video data, the method comprising:
determining a motion vector (mv);
determining a rounded motion vector being equivalent to a value of rmv in the following equations:
s=mv >>17;
f =Floor(Log 2(( mv{circumflex over ( )} s )|31))−4;
mask=(−1<< f )>>1;
round=(1<< f )>>2;
rmv =( mv +round)&mask; and
generating a motion vector predictor candidate based on the rounded motion vector.
2. The method of claim 1 , wherein the motion vector (mv) is derived as a motion vector corresponding to a collocated luma coding block inside a specified collocated picture.
3. The method of claim 1 , wherein the motion vector predictor candidate is generated based on the rounded motion vector by clipping a value within the range of −131072 to 131071.
4. A device comprising one or more processors configured to:
determine a motion vector (mv);
determine a rounded motion vector being equivalent to a value of rmv in the following equations:
s=mv >>17;
f =Floor(Log 2(( mv{circumflex over ( )} s )|31))−4;
mask=(−1<< f )>>1;
round=(1<< f )>>2;
rmv =( mv +round)&mask; and
generate a motion vector predictor candidate based on the rounded motion vector.