Deep Image Harmonization in Dual Color Spaces

TL;DR

This paper introduces a novel deep image harmonization method that leverages dual color spaces, combining RGB and Lab channels to improve the disentanglement of color and illumination features for more effective harmonization.

Contribution

It proposes a dual color space approach with a specialized network architecture that enhances feature disentanglement and control in image harmonization tasks.

Findings

Improved harmonization quality over RGB-only methods

Effective disentanglement of color and illumination features

Enhanced control over harmonization process

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 $RGB$ color space, leading to entangled features and limited representation ability. In contrast, decorrelated color space (e.g., $Lab$) has decorrelated channels that provide disentangled color and illumination statistics. In this paper, we explore image harmonization in dual color spaces, which supplements entangled $RGB$ features with disentangled $L$, $a$, $b$ features to alleviate the workload in harmonization process. The network comprises a $RGB$ harmonization backbone, an $Lab$ encoding module, and an $Lab$ control module. The backbone is a U-Net network translating composite image to harmonized image. Three encoders in $Lab$ encoding module extract three control codes independently from $L$, $a$, $b$ channels, which are used to manipulate the decoder features in harmonization backbone via $Lab$ control module. Our code and model are available at \href{https://github.com/bcmi/DucoNet-Image-Harmonization}{https://github.com/bcmi/DucoNet-Image-Harmonization}.