Scholarlink

<p>With the increasing prevalence of multi-view video, depth maps have emerged as crucial auxiliary information. When employing traditional lossy methods to encode depth maps, significant distortion may occur, impacting the quality of the virtual synthetic view. The depth lookup table (DLT) offers a lossless coding approach for depth maps, achieving up to 25% bitrate reduction for certain indoor scenes. However, DLT requires a larger allocation of video header information, such as sequence parameter set (SPS), with DLT consuming nearly 65.5% of the bits in SPS. This paper addresses the critical challenge of header overhead in 3D-HEVC depth coding. We propose a syntax-aware framework with two key innovations: (1) a hierarchical recursive delta encoding algorithm that exploits depth sparsity (≤ 6% utilization in standard sequences) to reduce DLT header bits by 91.43%, and (2) an inter-view predictive signaling architecture that slashes SPS/VPS overhead from 65.5% to 5.6% through table inheritance and adaptive recursion control. Our method, validated through testing on 3DV sequences, achieves state-of-the-art DLT compression efficiency while maintaining full backward compatibility with 3D-HEVC syntax.</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

Advanced depth map compression techniques for efficient 3D video communication

  • 期刊论文

  • 发表日期：

    2026年2月3日

查看全文
  • Yangang Cai,
  • Zhilei Ling,
  • Jingyuan Tang,
  • Songlin Sun

摘要

With the increasing prevalence of multi-view video, depth maps have emerged as crucial auxiliary information. When employing traditional lossy methods to encode depth maps, significant distortion may occur, impacting the quality of the virtual synthetic view. The depth lookup table (DLT) offers a lossless coding approach for depth maps, achieving up to 25% bitrate reduction for certain indoor scenes. However, DLT requires a larger allocation of video header information, such as sequence parameter set (SPS), with DLT consuming nearly 65.5% of the bits in SPS. This paper addresses the critical challenge of header overhead in 3D-HEVC depth coding. We propose a syntax-aware framework with two key innovations: (1) a hierarchical recursive delta encoding algorithm that exploits depth sparsity (≤ 6% utilization in standard sequences) to reduce DLT header bits by 91.43%, and (2) an inter-view predictive signaling architecture that slashes SPS/VPS overhead from 65.5% to 5.6% through table inheritance and adaptive recursion control. Our method, validated through testing on 3DV sequences, achieves state-of-the-art DLT compression efficiency while maintaining full backward compatibility with 3D-HEVC syntax.