Fast 3D-HEVC depth intra coding based on boundary continuity

Mei Juan Chen, Jie Ru Lin, Yu Chih Hsu, Yi Sheng Ciou, Chia Hung Yeh*, Min Hui Lin, Lih Jen Kau, Chuan Yu Chang

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

5 Citations (Scopus)


The encoding format of the 3D extension of high efficiency video coding (3D-HEVC) consists of a multiview color texture and an associated depth map. Because of the unique characteristics of the depth map, advanced coding techniques are designed for depth map coding at the expense of computational complexity. In this paper, fast algorithms are conceived to accelerate the intra coding time of the depth map based on boundary continuity. First, the proposed fast prediction unit (PU) mode decision reduces the number of conventional intra prediction modes based on calculating the total sum of squares (TSS) of the PU boundaries. Second, the proposed fast depth modeling mode (DMM) decision makes use of the variances of the boundary pixels to determine the execution of the DMM. Third, the proposed coding unit (CU) early termination algorithm decides whether to further split the current CU by utilizing the thresholds of the TSS and the rate-distortion cost (RD-cost). The experimental results show that the proposed algorithm provides better performance in terms of coding speed and bitrate than the algorithm in previous work. The coding time of the depth map is reduced by 56.08%, while the Bjontegaard delta bitrate (BD-BR) is only increased by 0.32% for the synthesis view.

Original languageEnglish
Article number9440389
Pages (from-to)79588-79599
Number of pages12
JournalIEEE Access
Publication statusPublished - 2021


  • 3D-HEVC
  • boundary continuity
  • depth map
  • fast algorithm
  • intra coding

ASJC Scopus subject areas

  • General Computer Science
  • General Materials Science
  • General Engineering


Dive into the research topics of 'Fast 3D-HEVC depth intra coding based on boundary continuity'. Together they form a unique fingerprint.

Cite this