Research of Chinese Painting Inpainting Based on Improved Criminisi Algorithm

Qiang Zhong, Bai Yu, Qing Wang, Qian Zhou, Mao Han

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

Abstract

To address the problems of incorrect inpainting order and incorrect similar patch matching when restoring Chinese paintings using the Criminisi algorithm, an inpainting algorithm with improved priority calculation and similar patch search matching is proposed. Firstly, edge features are extracted from the image, and gradients are calculated on the edge feature map to replace the data items in the original algorithm. The distance and the difference in the colour mean of the patch are introduced into the similar patch searching matching formula to ensure the continuity of the inpainting results. Experiments show that the proposed improved algorithm has better inpainting results on Chinese painting inpainting and better processing for texture details.

Original languageEnglish
Title of host publicationProceedings - 2022 2nd International Conference on Computer Graphics, Image and Virtualization, ICCGIV 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages217-222
Number of pages6
ISBN (Electronic)9781665492508
DOIs
StatePublished - 2022
Externally publishedYes
Event2nd International Conference on Computer Graphics, Image and Virtualization, ICCGIV 2022 - Chongqing, China
Duration: 23 Sep 202225 Sep 2022

Publication series

NameProceedings - 2022 2nd International Conference on Computer Graphics, Image and Virtualization, ICCGIV 2022

Conference

Conference2nd International Conference on Computer Graphics, Image and Virtualization, ICCGIV 2022
Country/TerritoryChina
CityChongqing
Period23/09/2225/09/22

Keywords

  • Chinese painting inpainting
  • component
  • Criminis algorithm
  • edge features
  • priority calculation

Fingerprint

Dive into the research topics of 'Research of Chinese Painting Inpainting Based on Improved Criminisi Algorithm'. Together they form a unique fingerprint.

Cite this