Time Complexity Reduction of CU Splitting in HEVC Intra Coding Using Zerocross Edge Detection Method
Md. Zahirul Islam, Tanvir Ahmed Redoy, Mansura Naznine, Mastura Khatun, Rizoan Toufiq, Mir Safwan Marzouq, Md Jawad Bin Khaleque · 2023
Video compression is a crucial technique for efficiently storing and transmitting videos over networks. The prevalent High Efficiency Video Coding (HEVC/H.265) standard achieves significant bitrate reduction but at the cost of increased time complexity. HEVC relies on coding tree unit (CTU) partitioning and Rate Distortion Optimization (RDO) during this process, which involves extensive rate-distortion trade-off searches. Determining the optimal partition for each Coding Unit (CU) within video frames poses a significant complexity challenge given the multitude of combinations to evaluate. As the number of CUs and partitioning grows, so does computational complexity. To address this, we propose an innovative heuristic model based on edge detection within HEVC’s intra-mode. Our model employs the Zerocross edge detection algorithm identifying edges in each frame by detecting zero crossings in the second derivative of the image. Experiments show that up to 49.49% reduction in encoding time for intra-mode HEVC across 12 commonly used test video sequences is possible when implementing our approach and omitting the RDO process. However, this efficiency gain comes with an average 3.35% bit-rate increase compared to HEVC.