SSH: A Self-Supervised Framework for Image Harmonization

TL;DR

This paper introduces SSH, a self-supervised framework for image harmonization that learns from natural images without annotations, improving compositing quality by matching appearance attributes.

Contribution

The paper presents a novel self-supervised approach for image harmonization that does not require annotated datasets, utilizing a dual data augmentation strategy and a new representation fusion perspective.

Findings

Outperforms state-of-the-art methods in metrics and visual quality

Uses a new dataset created by experts for benchmarking

Demonstrates effectiveness of self-supervised learning in image harmonization

Abstract

Image harmonization aims to improve the quality of image compositing by matching the "appearance" (\eg, color tone, brightness and contrast) between foreground and background images. However, collecting large-scale annotated datasets for this task requires complex professional retouching. Instead, we propose a novel Self-Supervised Harmonization framework (SSH) that can be trained using just "free" natural images without being edited. We reformulate the image harmonization problem from a representation fusion perspective, which separately processes the foreground and background examples, to address the background occlusion issue. This framework design allows for a dual data augmentation method, where diverse [foreground, background, pseudo GT] triplets can be generated by cropping an image with perturbations using 3D color lookup tables (LUTs). In addition, we build a real-world harmonization dataset as carefully created by expert users, for evaluation and benchmarking purposes. Our results show that the proposed self-supervised method outperforms previous state-of-the-art methods in terms of reference metrics, visual quality, and subject user study. Code and dataset are available at \url{https://github.com/VITA-Group/SSHarmonization}.