JVET-AP0253 AHG16: Comparative Study on RDO count and EncT under tool off test [Y. Tokumo, T. Ikai, S. Hong, K.-W. Liang (Sharp)] [late]
This contribution presents comparative study of the RDO count, calculated as the weighted sum of all RDOs, and EncT under tool-off testing using AHG16 software. The experiment indicates that, for some tools, the RDO count and EncT values are similar, while for others, they exhibit different behaviour. This suggests that using RDOs as well as EncT can be helpful for evaluating the complexity characteristics of each tool. It is recommended to use the AHG16 software metrics and consider their potential application in the reference software for the next standard development.
The tool-off tests were performed on a per-tool basis. There are cases where the number of RDOs increases, perhaps because more residual needs to be encoded. In some cases, the number of RDOs increases but encoding time decreases when a tool is switched off. This might indicate that number of RDOs might not be useful as a standalone criterion for the benefit of a tool.
It was concluded that none of the contributions in sections 4.8 and 4.15 provided new aspects to be included in the complexity reporting template – a new version JVET-AP2040 is not needed.