DiffCR: A Fast Conditional Diffusion Framework for Cloud Removal from Optical Satellite Images

TL;DR

DiffCR introduces a fast, diffusion-based framework for cloud removal in optical satellite images, outperforming existing methods in quality and efficiency by leveraging conditional guided diffusion and novel fusion blocks.

Contribution

The paper proposes a new diffusion model framework with a decoupled encoder and efficient fusion blocks for improved cloud removal in satellite imagery, with significantly reduced computational costs.

Findings

Achieves state-of-the-art performance on benchmark datasets.

Uses only about 5% of the parameters and computational resources of previous methods.

Demonstrates robust and high-quality cloud removal results.

Abstract

Optical satellite images are a critical data source; however, cloud cover often compromises their quality, hindering image applications and analysis. Consequently, effectively removing clouds from optical satellite images has emerged as a prominent research direction. While recent advancements in cloud removal primarily rely on generative adversarial networks, which may yield suboptimal image quality, diffusion models have demonstrated remarkable success in diverse image-generation tasks, showcasing their potential in addressing this challenge. This paper presents a novel framework called DiffCR, which leverages conditional guided diffusion with deep convolutional networks for high-performance cloud removal for optical satellite imagery. Specifically, we introduce a decoupled encoder for conditional image feature extraction, providing a robust color representation to ensure the close similarity of appearance information between the conditional input and the synthesized output. Moreover, we propose a novel and efficient time and condition fusion block within the cloud removal model to accurately simulate the correspondence between the appearance in the conditional image and the target image at a low computational cost. Extensive experimental evaluations on two commonly used benchmark datasets demonstrate that DiffCR consistently achieves state-of-the-art performance on all metrics, with parameter and computational complexities amounting to only 5.1% and 5.4%, respectively, of those previous best methods. The source code, pre-trained models, and all the experimental results will be publicly available at https://github.com/XavierJiezou/DiffCR upon the paper's acceptance of this work.