Back to Search Document details
12th Meeting: Macao, October 2018 2018-10-07 08:05
CE3-related: Classification-based mean value for CCLM coefficients derivation
Authors: X. Ma Huawei Fan Mu Huawei Hyeong-Gyun Yang Huawei Jie Chen Huawei
Abstract
In this contribution, a coefficients derivation method based on classification-based mean value is proposed. In the proposed method, the luma template samples are classified into two luma classes by the mean value of them. Correspondingly, two chroma classes are obtained. Then two luma mean values of the two luma classes, and two chroma mean values of the two chroma classes are obtained, respectively. CCLM coefficients are derived based on the two luma mean values and the two chroma mean values. Simulation results reportedly show BD rate on Y, Cb and Cr components for AI configuration over VTM2.0.1 is - 0.07%, 0.05%, and -0.44%, with 100% EncT, 98% DecT;
Meeting Notes
JVET-L0342 CE3-related: Classification-based mean value for CCLM coefficients derivation [X. Ma, F. Mu, H. Yang, J. Chen (Huawei)]

In this contribution, a coefficients derivation method based on classification-based mean value is proposed. In the proposed method, the luma template samples are classified into two luma classes by the mean value of them. Correspondingly, two chroma classes are obtained. Then two luma mean values of the two luma classes, and two chroma mean values of the two chroma classes are obtained, respectively. CCLM coefficients are derived based on the two luma mean values and the two chroma mean values. Simulation results reportedly show BD rate on Y, Cb and Cr components for AI configuration over VTM2.0.1 is - 0.07%, 0.05%, and -0.44%, with 100% EncT, 98% DecT;

Study in CE; requires complexity analysis in comparison to the simplified method that was adopted at this meeting. It likely requires more additions than JVET-L0191, and potentially two passes to determine the overall luma mean and the class means of luma and chroma. On the other hand, it is significantly less complex than previous CCLM and seems to perform better than JVET-L0191.

Cited:
JVET-L0023 JVET-L0651 X JVET-L1023 X JVET-M0093 JVET-M0857
ARCHIVE v4 :
JVET-L0342.pptx JVET-L0342_CCLM.xlsm JVET-L0342_CCLM_1row_1column_with_downsample.xlsm JVET-L0342_CCLM_1row_1column_without_downsample.xlsm JVET-L0342_r2.docx
Decisions
adopted
Study in CE; requires complexity analysis in comparison to the simplified method that was adopted at this meeting. It likely requires more additions than JVET-L0191, and potentially two passes to determine the overall luma mean and the class means of luma and chroma. On the other hand, it is significantly less complex than previous CCLM and seems to perform better than JVET-L0191.
Citation