Multi-Type Tree Partition Decision for H.266/VVC Inter Coding Based on Random Forest Classifiers

The flexible coding unit block partition structure of the quadtree with nested multi-type tree (MTT) for the newest video coding standard H.266/versatile video coding (VVC) can provide better coding efficiency than that of H.265/high efficiency video coding (HEVC). However, the computational complexity increases markedly. This paper proposes a fast MTT partition algorithm based on machine learning for the inter prediction of the H.266/VVC encoder. The proposed approach utilizes Random Forest classifiers to skip the least necessary MTT split. Experimental results show that the proposed algorithm reduces the encoding time with negligible Bjontegaard delta bitrate (BDBR) under the random access configuration. The proposed method outperforms the previous work.