High precision explicit weighted prediction for video coding

摘要

In one example, a device for coding video data includes a memory configured to store video data, and a video coder configured to code a value for a syntax element representative of whether a high bit depth is enabled for the video data, and when the value for the syntax element indicates that the high bit depth is enabled: code a value for a syntax element representative of the high bit depth for one or more parameters of the video data, code values for the parameters such that the values for the parameters are representative of bit depths that are based on the value for the syntax element representative of the high bit depth, and code the video data based at least in part on the values for the parameters.

主权项

1. A method of decoding video data, the method comprising:
decoding a value for a syntax element representative of whether weighted prediction having a signaled bit depth is enabled for video data; when the value for the syntax element indicates that the weighted prediction having the signaled bit depth is enabled:
decoding a value for a syntax element representative of the signaled bit depth for one or more parameters of the video data, wherein the one or more parameters comprise a luma_offset_l0[i] parameter or a luma_offset_l1[i] parameter;decoding values for the parameters such that the values for the parameters are representative of bit depths that are based on the value for the syntax element representative of the signaled bit depth; anddecoding the video data based at least in part on the values for the parameters, wherein decoding the video data comprises, when performing a weighted prediction process for luminance samples:
calculating a value for a variable o0 used in the weighted prediction process as being equal to a value for a luma_offset_l0[refIdxL0] parameter, without calculating (1<<(the bit depth minus 8)); andcalculating a value for a variable o1 used in the weighted prediction process as being equal to a value for a luma_offset_l1[refIdxL1] parameter, without calculating (1<<(the bit depth minus 8)).