ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2308.02813
25
17

Deep Image Harmonization in Dual Color Spaces

5 August 2023
L. Tan
Jiangtong Li
Li Niu
Liqing Zhang
ArXivPDFHTML
Abstract

Image harmonization is an essential step in image composition that adjusts the appearance of composite foreground to address the inconsistency between foreground and background. Existing methods primarily operate in correlated RGBRGBRGB color space, leading to entangled features and limited representation ability. In contrast, decorrelated color space (e.g., LabLabLab) has decorrelated channels that provide disentangled color and illumination statistics. In this paper, we explore image harmonization in dual color spaces, which supplements entangled RGBRGBRGB features with disentangled LLL, aaa, bbb features to alleviate the workload in harmonization process. The network comprises a RGBRGBRGB harmonization backbone, an LabLabLab encoding module, and an LabLabLab control module. The backbone is a U-Net network translating composite image to harmonized image. Three encoders in LabLabLab encoding module extract three control codes independently from LLL, aaa, bbb channels, which are used to manipulate the decoder features in harmonization backbone via LabLabLab control module. Our code and model are available at \href{https://github.com/bcmi/DucoNet-Image-Harmonization}{https://github.com/bcmi/DucoNet-Image-Harmonization}.

View on arXiv
Comments on this paper