Of itu-t sg16 wp3 and iso/iec jtc1/SC29/WG11

JCTVC-M0255 (m28712) Decoding a 10-bit HEVC sequence using an 8-bit decoder [D. Flynn, G. Martin-Cocher, D. He (RIM)]

HEVC, like other well-known codecs, contains profiles that accommodate the decoding of non-8-bit video sequences. In particular, HEVC contains a 10-bit profile that permits the carriage of 10-bit video sequences. However, for some implementations, the extra resources required for conformance to a higher bit-depth operating point can be prohibitive, yet there exist circumstances where a decoder may wish to offer best-effort decoding of a bitstream. Two techniques are presented that facilitate the decoding of 10-bit video bitstreams using an 8-bit decoder (but a 10-bit parser). The first uses the normal 8-bit reconstruction process, while the second involves addition of a rounding process during the reconstruction process. PSNR losses using low-delay main-10 bitstreams reportedly average at a 6 dB and 2.5 dB respectively. Further results are presented and specification text is provided for a possible annex.

Two example reduced-precision decoding processes were described.

Potential normative and informative approaches were discussed.

The normative approach could potentially specify a decoding profile that has no corresponding bitstream conformance profile, but rather uses decoding of a bitstream of a higher capability bitstream conformance profile.

The functionality seems useful, but a requirement would need to be established (at the parent body level) for a normative approach.

It was commented that there are other reduced-quality decoding schemes that get applied in practice (e.g. for trick mode playback), and these are generally not documented in the standard (at least not normatively). The presenter suggested that this process might differ more substantially from the ordinary decoding process than what might apply in other such cases.

The scheme was presented specifically for 8 bits decoding of 10 bit video, although it could apply to other combinations.

It seems desirable to study this type of scheme, consider what is the best way to perform such a reduced-precision decoding process, analyze the conditions under which it might work well and under which it would not work so well.

It was suggested that having some degree of encoder optimization to mitigate the effect might provide a benefit.

For further study.