Motion vector predictors (MVPs) for bi-predictive inter mode in video coding

A method of encoding video data in which adaptive motion vector prediction (AMVP) is used to code a first motion vector and a second motion vector associated with a current video block in accordance with a bi-predictive inter mode that is based on a first reference block and a second reference video block, the method comprising:

• determining the first motion vector to be used to code the current video block in the bi-predictive inter mode, wherein the first motion vector is a function of the current video block and the first reference video block;
• identifying a first motion vector predictor (MVP) based on a first set of motion vector predictors;
• generating a first index that identifies the first MVP;
• generating a first motion vector difference (MVD) based on the first motion vector and the first MVP;
• determining the second motion vector to be used to code the current video block in the bi-predictive inter mode;
• identifying a second MVP based on a second set of motion vector predictors, wherein the second set of motion vector predictors includes a candidate motion vector predictor based on a scaled first motion vector, wherein the scaled first motion vector is formed by scaling the first motion vector by a scale factor that is function of a temporal distance from the current video block to the second reference video block and a temporal distance from the current video block to the first reference video block;
• generating a second index that identifies the candidate motion vector predictor as the second MVP;
• generating a second MVD based on the second motion vector and the second MVP;
• and
• outputting the first index, the first MVD, the second index and the second MVD as part of an encoded bitstream for the current video block.

A video decoding device that decodes video data encoded in a bi-predictive inter mode that is based on a first reference block and a second reference video block, the video decoding device comprising:

• a memory;
• and
• a video decoder, connected to the memory, wherein the video decoder is configured to:
• receive a first motion vector difference (MVD) associated with the first motion vector;
• receive a first index that identifies a first motion vector predictor (MVP), selected from a first set of motion vector predictors;
• generate the first motion vector based on the first MVD and the first MVP, the first motion vector a function of the current video block and the first reference video block;
• receive a second MVD associated with the second motion vector;
• receive a second index that identifies a second MVP, wherein the second MVP is selected from a second set of motion vector predictors, the second set of motion vector predictors including a candidate motion vector predictor that is a function of a scaled first motion vector found by scaling the first motion vector by a scale factor that is function of a temporal distance from the current video block to the second reference video block and a temporal distance from the current video block to the first reference video block;
• and
• generate the second motion vector based on the second MVD and the candidate motion vector predictor.

A video encoding device that encodes video data in a current video block in a bi-predictive inter mode that is based on a first reference block and a second reference video block, the video encoding device comprising:

• a memory;
• and
• a video encoder connected to the memory, wherein the video encoder is configured to:
• determine a first motion vector to be used to code the current video block in the bi-predictive inter mode, wherein the first motion vector is based on the current video block and the first reference video block;
• identify a first motion vector predictor (MVP) based on a first set of motion vector predictors;
• generate a first index that identifies the first MVP;
• generate a first motion vector difference (MVD) based on the first motion vector and the first MVP;
• determine a second motion vector to be used to code the current video block in the bi-predictive inter mode, wherein the second motion vector is based on the current video block and the second reference video block;
• identify a second MVP based on a second set of motion vector predictors, wherein the second set of motion vector predictors includes a candidate motion vector predictor based on a scaled first motion vector, wherein the scaled first motion vector is formed by scaling the first motion vector by a scale factor that is function of a temporal distance from the current video block to the second reference video block and a temporal distance from the current video block to the first reference video block;
• generate a second index that identifies the candidate motion vector predictor as the second MVP;
• generate a second MVD based on the second motion vector and the second MVP;
• and
• output the first index, the first MVD, the second index and the second MVD as part of an encoded bitstream for the current video block.

A device for decoding video data in which adaptive motion vector prediction (AMVP) is used to code a first motion vector and a second motion vector associated with a current video block that is coded in a bi-predictive inter mode that is based on a first reference block and a second reference video block, the device comprising:

• means for receiving a first motion vector difference (MVD) associated with the first motion vector;
• means for receiving a first index that identifies a first motion vector predictor (MVP), the first MVP selected from a first set of motion vector predictors;
• means for generating the first motion vector based on the first MVD and the first MVP, the first motion vector a function of the current video block and the first reference video block;
• means for receiving a second MVD associated with the second motion vector;
• means for receiving a second index that identifies a second MVP, wherein the second MVP is selected from a second set of motion vector predictors, the second set of vectors motion vector predictors including a candidate motion vector predictor that is a function of a scaled first motion vector found by scaling the first motion vector by a scale factor that is function of a temporal distance from the current video block to the second reference video block and a temporal distance from the current video block to the first reference video block;
• and
• means for generating the second motion vector based on the second MVD and the candidate motion vector predictor.

A device for encoding video data in which adaptive motion vector prediction (AMVP) is used to code a first motion vector and a second motion vector associated with a current video block in accordance with a bi-predictive inter mode that is based on a first reference block and a second reference video block, the device comprising:

• means for determining the first motion vector to be used to code the current video block in the bi-predictive inter mode, wherein the first motion vector is a function of the current video block and the first reference video block;
• means for identifying a first motion vector predictor (MVP) based on a first set of motion vector predictors;
• means for generating a first index that identifies the first MVP;
• means for generating a first motion vector difference (MVD) based on the first motion vector and the first MVP;
• means for determining the second motion vector to be used to code the current video block in the bi-predictive inter mode;
• means for identifying a second MVP based on a second set of motion vector predictors, wherein the second set of motion vector predictors includes a candidate motion vector predictor based on a scaled first motion vector, wherein the scaled first motion vector is formed by scaling the first motion vector by a scale factor that is function of a temporal distance from the current video block to the second reference video block and a temporal distance from the current video block to the first reference video block;
• means for generating a second index that identifies the candidate motion vector predictor as the second MVP;
• means for generating a second MVD based on the second motion vector and the second MVP;
• and
• means for outputting the first index, the first MVD, the second index and the second MVD as part of an encoded bitstream for the current video block.

A non-transitory computer-readable medium comprising instructions that upon execution cause a processor to decode video data in which adaptive motion vector prediction (AMVP) is used to code a first motion vector and a second motion vector associated with a current video block that is coded in a bi-predictive inter mode that is based on a first reference block and a second reference video block, wherein the instructions cause the processor to:

• upon receiving a first motion vector difference (MVD) associated with the first motion vector, and upon receiving a first index that identifies a first motion vector predictor (MVP), the first MVP selected from a first set of motion vector predictors, generate the first motion vector based on the first MVD and the first MVP, the first motion vector associated with the current video block and the first reference video block;
• and
• upon receiving a second MVD associated with the second motion vector, and upon receiving a second index that identifies a second MVP, wherein the second MVP is selected from a second set of motion vector predictors, the second set of motion vector predictors including a candidate motion vector predictor that is a function of a scaled first motion vector found by scaling the first motion vector by a scale factor that is function of a temporal distance from the current video block to the second reference video block and a temporal distance from the current video block to the first reference video block, generate the second motion vector based on the second MVD and the candidate motion vector predictor.

A non-transitory computer-readable medium comprising instructions that upon execution cause a processor to encode video data in which adaptive motion vector prediction (AMVP) is used to code a first motion vector and a second motion vector associated with a current video block in accordance with a bi-predictive inter mode that is based on a first reference block and a second reference video block, wherein the instructions cause the processor to:

• determine the first motion vector to be used to code the current video block in the bi-predictive inter mode, wherein the first motion vector is a function of the current video block and the first reference video block;
• identify a first motion vector predictor (MVP) based on a first set of motion vector predictors;
• generate a first index that identifies the first MVP;
• generate a first motion vector difference (MVD) based on the first motion vector and the first MVP;
• determine the second motion vector to be used to code the current video block in the bi-predictive inter mode;
• identify a second MVP based on a second set of motion vector predictors, wherein the second set of motion vector predictors includes a candidate motion vector predictor based on a scaled first motion vector, wherein the scaled first motion vector is formed by scaling the first motion vector by a scale factor that is function of a temporal distance from the current video block to the second reference video block and a temporal distance from the current video block to the first reference video block;
• generate a second index that identifies the candidate motion vector predictor as the second MVP;
• generate a second MVD based on the second motion vector and the second MVP;
• and
• output the first index, the first MVD, the second index and the second MVD as part of an encoded bitstream for the current video block.