Region-based poisson blending for image repairing

Wei Cheng Chen, Wen-Jiin Tsai*

*Corresponding author for this work

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


Image inpainting is a widely used technique for image repairing, which utilizes information from the undamaged regions in the same image to do the inpainting. However, the repairing methods using inpainting cannot work well when the textures in the damaged area cannot be found in the remaining area of the image. In this paper, an image repairing method is presented, which uses Affine transformation and Poisson blending for repairing, based on the assumption that reference images are available. The reference image is first affine-transformed to have the same viewing angle and scaling size to the damaged image and then Poisson blending is applied to repair the damaged area by compositing the corresponding area in the affined reference image. Since the blending will preserve reference image’s textures and adjust its illumination and color tone according to the damaged image, the composition is supposed to be seamlessly. However, it was observed that using Poisson blending in image repairing may cause dramatically blurring in the resulting image sometimes. This mainly comes from the big differences in pixel-gradient changes in the blending area. To cope with the problem, a region-based Poisson blending is proposed, in which the damaged area is not blended as a whole, but segmented into regions and applied with blending separately. The experimental result shows that the blurring artifacts can be reduced obviously by using the proposed region-based approach.

Original languageEnglish
Title of host publicationIntelligent Systems and Applications - Proceedings of the 2018 Intelligent Systems Conference IntelliSys Volume 1
EditorsKohei Arai, Supriya Kapoor, Rahul Bhatia
PublisherSpringer Verlag
Number of pages15
ISBN (Print)9783030010539
StatePublished - 9 Nov 2018
EventIntelligent Systems Conference, IntelliSys 2018 - London, United Kingdom
Duration: 6 Sep 20187 Sep 2018

Publication series

NameAdvances in Intelligent Systems and Computing
ISSN (Print)2194-5357


ConferenceIntelligent Systems Conference, IntelliSys 2018
Country/TerritoryUnited Kingdom


  • Image repairing
  • Poisson blending
  • Region-based


Dive into the research topics of 'Region-based poisson blending for image repairing'. Together they form a unique fingerprint.

Cite this