Painterly Image Harmonization via Bi-Transformation with Dynamic Kernels


Zhangliang Sun (Tsinghua University, Tsinghua University), Hui Zhang (Tsinghua University)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)Supplementary

Abstract

Painterly image harmonization aims to achieve seamless integration of a foreground photographic object and a background painting by matching their complex artistic styles. Previous methods effectively retain foreground content but often fall short of adequately matching the styles such as strokes and patterns between the foreground and background. In this work, we explore the integration of dynamic networks for painterly image harmonization. We propose a novel bi-transformation model that utilizes dynamic kernels in a two-stage process. Through transformations based on dynamic kernels, the foreground and background feature maps can be well aligned to generate a harmonized image. Extensive experiments show that our model can achieve a more harmonized style fusion while retaining finer content.

Citation

@inproceedings{Sun_2024_BMVC,
author    = {Zhangliang Sun and Hui Zhang},
title     = {Painterly Image Harmonization via Bi-Transformation with Dynamic Kernels},
booktitle = {35th British Machine Vision Conference 2024, {BMVC} 2024, Glasgow, UK, November 25-28, 2024},
publisher = {BMVA},
year      = {2024},
url       = {https://papers.bmvc2024.org/0100.pdf}
}


Copyright © 2024 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection