JVET-Q0111 A proposal of an 8-bit profile [L. Zhang, Y.-K. Wang, J. Wang (Bytedance)]

This contribution was discussed in the joint meeting with the parent bodies on Tuesday 14 January. See the notes in section 11.3.

JVET-Q0111 A proposal of an 8-bit profile [L. Zhang, Y.-K. Wang, J. Wang (Bytedance), X. Li (Tencent), X. Wang (Kwai)]

In the contribution, it is asserted that hardware decoders of VVC take time to implement, while it would be desirable to deploy VVC as early as possible, initially using software decoders. However, the implementation cost of software decoders for 10-bit decoding is asserted to be significantly higher than that for 8-bit decoding due to the need of SIMD optimizations etc. This contribution thus proposes to specify an 8-bit profile for VVC.

Minimizing the time to initial deployment of the standard was mentioned as a significant factor in desiring an 8 bit profile. The proposal is intended to address the temporary need for near-term deployment.

The contributor asserted that some applications do not need to use 10 bit video.

It was asked what is the difference in terms of complexity or power consumption? 20-30% additional decoder runtime was estimated by a participant for running a 10-bit decoder when the video is 8 bit. If the decoder would be optimized to run differently for 8 and 10 bit video, there would not be an impact.

It was commented that this could pose some interop difficulties once 10 bit encoding becomes prevalent, as 8-bit decoders would not be able to decode the 10 bit encoded video.

It was noted that there is a coding efficiency benefit for encoding 8 bit source video using 10 bit encoding (perhaps 2–3%). This would involve a runtime increase.

A participant commented that an externally specified “sub-profile” could address the asserted temporary market need.

Insufficient support was expressed for this proposal, so no action was taken on it.