Back to Search Document details
12th Meeting: Macao, October 2018 2018-10-07 05:16
CE3-related: Simplification of PDPC
Abstract
This contribution presents modification of position dependent intra prediction combination (PDPC). It includes three modifications which are block size restriction, removal of interpolation, and restriction on wide angle intra prediction modes. It is reported that the average luma BD-rate gain of all the modifications for AI case is -0.03% compared to VTM 2.0.1.
JVET-L0152 CE3-related: Simplification of PDPC [J. Lee, H. Lee, S.-C. Lim, J. Kang, H. Y. Kim (ETRI)]

This contribution presents modification of position dependent intra prediction combination (PDPC). It includes three modifications which are block size restriction, removal of interpolation, and restriction on wide angle intra prediction modes. It is reported that the average luma BD-rate gain of all the modifications for AI case is -0.03% compared to VTM 2.0.1.

  • PDPC not used for64x64
  • PDPC applied to less prediction samples by modyfying the weight
  • no bilinear interpolation in PDPC
  • no PDPC for wide angle modes

The first two aspects give marginal gain

The third aspect would be a straightforward complexity reduction with marginal loss. It is however asked why PDPC is not using the nearest neighbour

The fourth aspect is not a complexity reduction, as PDPC needs to be implemented anyway for other modes.

Further study was encouraged on the third aspect.

PATENTS:
CN111699682B 0.34 2025-06-06 EP3298807B1 0.32 2023-08-30 US12425622B2 0.30 2025-09-23 EP3354091B1 0.28 2019-10-23 EP2848061B1 0.26 2016-03-16 EP3796710B1 0.24 2025-12-24 EP3472960B1 0.22 2024-12-11 EP3295750B1 0.20 2020-07-29
Decisions
Further study was encouraged on the third aspect.
Citation