Intra prediction for 360-degree video

A device for coding 360-degree video data, the device comprising

• a memory configured to store the 360-degree video data;
• and
• one or more processors implemented in circuitry, the one or more processors configured to:
• determine reference samples based on a location of a current block of a current picture of the 360-degree video data and a packing arrangement that defines an arrangement of a plurality of regions into the current picture, wherein:
• the current picture is in a projected domain,
• each respective region of the plurality of regions is a respective face defined by a projection of 360-degree video data, the projection being a cubemap projection (CMP) or a derivative projection of the CMP, the regions including a first region and a second region,
• the regions are arranged in the current picture according to the packing arrangement,
• based on the location of the current block being at a border of the first region that is adjacent to the second region and there being a discontinuity at the border due to the packing arrangement, the reference samples are samples of the current picture that spatially neighbor the current block in a spherical domain and not in the projected domain, and
• the one or more processors are configured such that, based on the current block being located at the border of the first region, the one or more processors:
• apply geometric derivation to determine a reference sample location;
• and
• determine a value of a reference sample, wherein:
• based on the reference sample location not being within any available area, the reference sample is a copy of a neighboring sample that neighbors the current sample in the spherical domain, and
• based on the reference sample location being within an available area, the reference sample is a sample at the reference sample location;
• and
• generate a prediction block for the current block using intra prediction based on one or more of the determined reference samples.

A device for coding video data, the device comprising:

• means for determining reference samples based on a location of a current block of a current picture of the 360-degree video data and a packing arrangement that defines an arrangement of a plurality of regions in the current picture, wherein:
• the current picture is in a projected domain,
• each respective region of the plurality of regions is a respective face defined by a projection of 360-degree video data, the projection being a cubemap projection (CMP) or a derivative projection of the CMP, the regions including a first region and a second region,
• the regions are arranged in the current picture according to the packing arrangement,
• based on the location of the current block being at a border of the first region that is adjacent to the second region and there being a discontinuity at the border due to the packing arrangement, the reference samples are samples of the current picture that spatially neighbor the current block in a spherical domain and not in the projected domain, and
• the means for determining the reference samples comprises, based on the current block being located at the border of the first region:
• means for applying geometric derivation to determine a reference sample location;
• and
• means for determining a value of a reference sample, wherein:
• based on the reference sample location not being within any available area, the reference sample is a copy of a neighboring sample that neighbors the current sample in the spherical domain, and
• based on the reference sample location being within an available area, the reference sample is a sample at the reference sample location;
• and
• means for generating a prediction block for the current block using intra prediction based on one or more of the determined reference samples.

A non-transitory computer-readable storage medium having stored thereon instructions that, when executed, cause one or more processors to:

• determine reference samples based on a location of a current block of a current picture of the 360-degree video data and a packing arrangement that defines an arrangement of a plurality of regions into the current picture, wherein:
• the current picture is in a projected domain,
• each respective region of the plurality of regions is a respective face defined by a projection of 360-degree video data, the projection being a cubemap projection (CMP) or a derivative projection of the CMP, the regions including a first region and a second region,
• the regions are arranged in the current picture according to the packing arrangement,
• based on the location of the current block being at a border of the first region that is adjacent to the second region and there being a discontinuity at the border due to the packing arrangement, the reference samples are samples of the current picture that spatially neighbor the current block in a spherical domain and not in the projected domain, and
• based on the current block is located at the border of the first region, execution of the instructions causes the one or more processors to:
• apply geometric derivation to determine a reference sample location;
• and
• determine a value of a reference sample, wherein:
• based on the reference sample location not being within any available area, the reference sample is a copy of a neighboring sample that neighbors the current sample in the spherical domain, and
• based on the reference sample location being within an available area, the reference sample is a sample at the reference sample location;
• and
• generate a prediction block for the current block using intra prediction based on one or more of the determined reference samples.