Efficient Intra Prediction in Versatile Video Coding Using Adaptive Partitioning and Structural Analysis

Versatile Video Coding (VVC) implements numerous advanced tools for improving compression efficiency and perceptual video quality at the expense of a substantial increase in computing complexity. One such tool is intra-frame coding. To resolve this computational problem, this paper proposes an efficient intra-coding approach using texture and structural analysis. The initial texture characteristics are derived through the early termination of coding unit (CU) splitting using the Scharr operator, which captures directional gradient information. Next, the extraction of structural features is based on a dissimilarity-driven criterion to skip redundant ternary tree (TT) splits and thus facilitate early termination of ternary CU partitioning. Compared with state-of-the- art methods, the proposed method yields a 45.79% decrease in encoding time with a minimal increase of 1.23% in BD-bitrate across standard test sequences.