Sub-PU level advanced residual prediction

A method of encoding video data, the method comprising:

• splitting a prediction unit (PU) of a coding unit (CU) into two or more sub-PUs including a first sub-PU and a second sub-PU, such that the two or more sub-PUs of the PU are non-overlapping portions of the PU;
• obtaining a first motion vector of a first type for the first sub-PU and a second motion vector of the first type for the second sub-PU;
• obtaining a third motion vector of a second type for the first sub-PU and a fourth motion vector of the second type for the second sub-PU, such that the second type is different than the first type;
• encoding a first portion of the CU corresponding to the first sub-PU according to advanced residual prediction (ARP) using the first motion vector and the third motion vector;
• and
• encoding a second portion of the CU corresponding to the second sub-PU according to ARP using the second motion vector and the fourth motion vector.

A device for coding video data, the device comprising:

• a memory configured to store video data;
• and
• one or more processors coupled the memory and configured to:
• split a prediction unit (PU) of a coding unit (CU) into two or more sub-PUs including a first sub-PU and a second sub-PU, such that the two or more sub-PUs of the PU are non-overlapping portions of the PU;
• obtain a first motion vector of a first type for the first sub-PU and a second motion vector of the first type for the second sub-PU;
• obtain a third motion vector of a second type for the first sub-PU and a fourth motion vector of the second type for the second sub-PU, such that the second type is different than the first type;
• code a first portion of the CU corresponding to the first sub-PU according to advanced residual prediction (ARP) using the first motion vector and the third motion vector;
• and
• code a second portion of the CU corresponding to the second sub-PU according to ARP using the second motion vector and the fourth motion vector.

A non-transitory computer-readable storage medium having stored thereon instructions that, when executed, cause at least one processor to:

• split a prediction unit (PU) of a coding unit (CU) into two or more sub-PUs including a first sub-PU and a second sub-PU, such that the two or more sub-PUs of the PU are non-overlapping portions of the PU;
• obtain a first motion vector of a first type for the first sub-PU and a second motion vector of the first type for the second sub-PU;
• obtain a third motion vector of a second type for the first sub-PU and a fourth motion vector of the second type for the second sub-PU, such that the second type is different than the first type;
• code a first portion of the CU corresponding to the first sub-PU according to advanced residual prediction (ARP) using the first motion vector and the third motion vector;
• and
• code a second portion of the CU corresponding to the second sub-PU according to ARP using the second motion vector and the fourth motion vector.