Acta Optica Sinica, Volume. 45, Issue 17, 1720016(2025)

Portrait Content Generation Technologies for 3D Light Field Displays (Invited)

Sheng Shen, Xinzhu Sang*, Shujun Xin, and Binbin Yan
Author Affiliations
  • School of Electronic Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • show less
    Figures & Tables(14)
    3D Portrait reconstruction methods
    Light Stage device[35]
    Images captured by Light Stage[35]
    Facial reflectance function[35]
    Sampling of facial reflectance function[35]. (a) Frontal region; (b) right side of chin; (c) junction between cheek and base of the nose; (d) inner part of auricle
    Compute relighting pixel values[35]
    Face rendered under sampled illumination[35]
    Generating highly realistic relighting results under various lighting conditions and viewing angles[5]
    Results of 3D light field display under different lighting conditions[5]
    Network model flowchart
    3D light field display effects under different facial expressions
    Overall flowchart of real-time reconstruction-driven portrait
    Light field display results based on eye-tracking[63]. (a) Public dataset; (b) self-built dataset
    • Table 1. Comparison of typical 3D portrait reconstruction algorithms

      View table

      Table 1. Comparison of typical 3D portrait reconstruction algorithms

      YearMethodProsConsDatasetPerformance
      2022HeadNeRF[15]Good reconstruction performance; capable of hair reconstructionSlow reconstruction speedFFHQPSNR is 24.90
      2023HRN[31]Portrait reconstruction supports single or multiple viewsPoor reconstruction quality for occluded parts in collected dataFaceScapeNMSE is 0.065
      2024VolTeMorph[27]Reconstructs and generates photo-realistic portrait scenes; supports real-time renderingPoor reconstruction on mouth and eyesD-NeRFPSNR is 30.20
      2025HRAvatar[32]Reconstructs high-fidelity re-illuminated portraits from monocular videosPoor hair and accessory reconstructionINSTA, self-capturedPSNR is 30.36, PSNR is 28.97
    Tools

    Get Citation

    Copy Citation Text

    Sheng Shen, Xinzhu Sang, Shujun Xin, Binbin Yan. Portrait Content Generation Technologies for 3D Light Field Displays (Invited)[J]. Acta Optica Sinica, 2025, 45(17): 1720016

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Optics in Computing

    Received: Jun. 4, 2025

    Accepted: Aug. 14, 2025

    Published Online: Sep. 3, 2025

    The Author Email: Xinzhu Sang (xzsang@bupt.edu.cn)

    DOI:10.3788/AOS251216

    CSTR:32393.14.AOS251216

    Topics