Transform Selection for Joint Coding of Chroma Residuals

초록

The work for beyond VVC (Versatile Video Coding) capability in JVET aims to surpass VVC in coding efficiency by using various efficient tools such as advanced transforms. While the luma channel benefits from multiple transform selection (MTS), the chroma channels remain restricted to the default DCT-II pair. This limitation is particularly undesirable for joint coding of chroma residuals (JCCR). Even after JCCR reduces inter-component redundancy by combining residuals, the generated joint residuals exhibit diverse spatial characteristics that the fixed DCT-II pair cannot efficiently decorrelate. To address this limitation, this paper proposes a transform selection method for JCCR-coded blocks. Instead of constructing a new transform set, the proposed method reuses the explicit MTS set defined in luma. Furthermore, the proposed method is applied exclusively to cross-component prediction (CCP) modes to maximize coding efficiency. Experimental results on ECM-17.0 show BDBR coding gains of 0.01% (Y), 0.12% (Cb), with a 0.01% degradation (Cr).

키워드