Context reduction for context adaptive binary arithmetic coding

摘要

A reduction in the number of binarizations and/or contexts used in context adaptive binary arithmetic coding (CABAC) for video coding is proposed. In particular, this disclosure proposes techniques that may lower the number contexts used in CABAC by up to 56.

主权项

1. A method of encoding video data, the method comprising:
determining a first prediction type for a block of video data in a P slice; representing the first prediction type as a P slice prediction type syntax element; determining a second prediction type for a block of video data in a B slice; representing the second prediction type as a B slice prediction type syntax element; determining a third prediction type for a block of video data in an I slice; representing the third prediction type as an I slice prediction type syntax element; determining a P slice binarization for the P slice prediction type syntax element using a first binarization mapping; determining a B slice binarization for the B slice prediction type syntax element using the first binarization mapping; determining an I slice binarization for the I slice prediction type syntax element using a second binarization mapping, the second binarization mapping being different than the first binarization mapping; and encoding the video data based on the binarizations of the P slice prediction type syntax element, the B slice prediction type syntax element, and the I slice prediction type syntax element, wherein each of the first prediction type, second prediction type, and third prediction type specifies a respective prediction mode and a respective partition type.