Scholarlink

<p>Video compression inherently makes trade-off among among bit rate, distortion and computational complexity. Exhaustive rate-distortion (R–D) optimized mode decision including block partitioning and skip mode decision imposes prohibitive computational complexity in modern codecs like the High Efficiency Video Coding (HEVC) and the Versatile Video Coding (VVC). To reduce computational complexity, early skip decision (ESD) techniques are employed to terminate block partitioning before full exploration. However, traditional ESD methods rely primarily on objective quality metrics (e.g., PSNR), which often fail to align with human perceptual quality, potentially sacrificing visual fidelity for encoding speed. Although incorporating perceptual metrics into encoder in-loop R-D optimization addresses this limitation, this approach inevitably increases computational complexity significantly. To address the problems, we propose a perceptual quality oriented R-D optimization video coding framework based on interaction between pre-encoding processing and in-loop R-D optimized mode decision. Firstly, we introduce a perceptual distortion based preprocessing algorithm, which exploits simplified visual information fidelity (SVIF). The SVIF-based preprocessing (SVIFPP) algorithm eliminates perceptual redundancy through a SVIF-based R-D optimization model. By selectively replacing blocks in the current frame with blocks in reconstructed reference frames, we prioritize regions where the block replacement minimizes perceptual distortion. Consequently, SVIFPP generates preprocessed sequence with selective region-replacement and masks indicating such regions. The masks are intrinsically related to skip mode that efficiently guides the R-D optimized mode decision in a perceptual manner. Secondly, we propose a mask-guided early skip decision (MGESD) mechanism that uses the masks generated by preprocessing to terminate further block partitioning during encoding process, bypassing redundant R-D optimized mode decision. Experiments on the HEVC and VVC test models demonstrate that our proposed framework achieves significant encoding acceleration, reducing average encoding time by 17.1% for HEVC and 16.0% for VVC. It also delivers improved R-D performance, achieving BD-rates of −0.2% on HEVC and −0.1% on VVC. The synergy of preprocessing and mask-guided encoding establishes a novel paradigm for complexity reduction in perceptual video coding.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A fast video coding scheme based on perceptual rate distortion optimized preprocessing

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • Luheng Jia,
  • Yifan Zang,
  • Jiyong Yu,
  • Zhaoyang Yu,
  • Shuyuan Zhu,
  • Li Song,
  • Kebin Jia

摘要

Video compression inherently makes trade-off among among bit rate, distortion and computational complexity. Exhaustive rate-distortion (R–D) optimized mode decision including block partitioning and skip mode decision imposes prohibitive computational complexity in modern codecs like the High Efficiency Video Coding (HEVC) and the Versatile Video Coding (VVC). To reduce computational complexity, early skip decision (ESD) techniques are employed to terminate block partitioning before full exploration. However, traditional ESD methods rely primarily on objective quality metrics (e.g., PSNR), which often fail to align with human perceptual quality, potentially sacrificing visual fidelity for encoding speed. Although incorporating perceptual metrics into encoder in-loop R-D optimization addresses this limitation, this approach inevitably increases computational complexity significantly. To address the problems, we propose a perceptual quality oriented R-D optimization video coding framework based on interaction between pre-encoding processing and in-loop R-D optimized mode decision. Firstly, we introduce a perceptual distortion based preprocessing algorithm, which exploits simplified visual information fidelity (SVIF). The SVIF-based preprocessing (SVIFPP) algorithm eliminates perceptual redundancy through a SVIF-based R-D optimization model. By selectively replacing blocks in the current frame with blocks in reconstructed reference frames, we prioritize regions where the block replacement minimizes perceptual distortion. Consequently, SVIFPP generates preprocessed sequence with selective region-replacement and masks indicating such regions. The masks are intrinsically related to skip mode that efficiently guides the R-D optimized mode decision in a perceptual manner. Secondly, we propose a mask-guided early skip decision (MGESD) mechanism that uses the masks generated by preprocessing to terminate further block partitioning during encoding process, bypassing redundant R-D optimized mode decision. Experiments on the HEVC and VVC test models demonstrate that our proposed framework achieves significant encoding acceleration, reducing average encoding time by 17.1% for HEVC and 16.0% for VVC. It also delivers improved R-D performance, achieving BD-rates of −0.2% on HEVC and −0.1% on VVC. The synergy of preprocessing and mask-guided encoding establishes a novel paradigm for complexity reduction in perceptual video coding.