Repairing IR depth image with 2D RGB image

Yeh Wei Yu, Tzu Kai Wang, Chi Chung Lau, Jia Ching Wang, Tsung Hsun Yang, Jann Long Chern, Ching Cherng Sun

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


3D-image technologies produce RGBD images use 2D RGB image combing with depth image. The most popular depth imaging methods including structure light, time-of-flight, and stereoscopic image. Since no point cloud detection method can prevent depth information loss, the complement methods is therefore important. Therefore, we proposed depth information complement method, which use a novel iterative low-pass pervasion method to apply depth image pervasion. We make two experiments to demonstrate the idea. The first experiment induces strong IR noise to destroy parts of the depth image. Then we successfully complement the depth image by the proposed method. The second experiment demonstrates the complement performance of iterative low-pass pervasion method and shows good Accuracy and Precision.

Original languageEnglish
Title of host publicationCurrent Developments in Lens Design and Optical Engineering XIX
EditorsR. Barry Johnson, Simon Thibault, Virendra N. Mahajan
ISBN (Electronic)9781510620612
StatePublished - 2018
Event19th Conference on Current Developments in Lens Design and Optical Engineering 2018 - San Diego, United States
Duration: 21 Aug 201822 Aug 2018

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X


Conference19th Conference on Current Developments in Lens Design and Optical Engineering 2018
Country/TerritoryUnited States
CitySan Diego


  • 3D-image
  • Holistically-Nested Edge Detection method
  • IR noise
  • RGB image
  • RGBD image
  • depth image complementation
  • depth information loss
  • iterative low-pass pervasion


Dive into the research topics of 'Repairing IR depth image with 2D RGB image'. Together they form a unique fingerprint.

Cite this