Back to Search Document details
16th Meeting: Geneva, October 2019 2019-09-25 04:06
CE3-related: CCLM with unified filter shape
Abstract
At the previous Gothenburg meeting, JVET-O0343 proposed a CCLM simplification method in both H/W and S/W implementation-wise. In JVET-O0343, unavailable neighbouring samples are substituted by available sample values using reference sample construction process, so that it is able to reduce conditional branches and unify down-sampling filter shapes. In this contribution, the complexity analysis on the additional operation by the reference sample construction process and the number of down-sampling filter shapes to be used are provided to compare the current CCLM with the proposed CCLM. The proposed CCLM is implemented on top of VTM-6.0 and shows 0.01%/0.03%/0.06% and x.xx%/x.xx%/x.xx% BD-rate impact on Y/Cb/Cr components for AI and RA cases, respectively.
JVET-P0265 CE3-related: CCLM with unified filter shape [D.-Y. Kim (Chips&Media), S.-C. Lim, J. Lee, J. Kang, H. Lee (ETRI)]

At the previous Gothenburg meeting, JVET-O0343 proposed a CCLM simplification method in both H/W and S/W implementation-wise. In JVET-O0343, unavailable neighbouring samples are substituted by available sample values using reference sample construction process, so that it is able to reduce conditional branches and unify down-sampling filter shapes. In this contribution, the complexity analysis on the additional operation by the reference sample construction process and the number of down-sampling filter shapes to be used are provided to compare the current CCLM with the proposed CCLM. The proposed CCLM is implemented on top of VTM-6.0 and shows 0.01%/0.03%/0.06% and 0.01%/0.13%/0.14% BD-rate impact on Y/Cb/Cr components for AI and RA cases, respectively.

It is asked whether luma padding is needed in dual tree case for chroma CCLM. This is not needed in this case. In single tree, how many lines need to be padded for luma? One line only. In the proposed method 3 luma lines need to be padded. It was commented that the proposed method is not a simplification. There is also some coding loss reported. Last meeting the decision was no action and no CE test. What is different in the current proposal? There is no difference, only some complexity analysis.

It is also commented that removing filters that were designed for other chroma sampling formats is not desirable. No results are provided.

The BoG recommended no action.

Decisions
The BoG recommended no action.
Citation