RSDiff: Remote Sensing Image Generation from Text Using Diffusion Model

TL;DR

This paper presents RSDiff, a two-stage diffusion model that synthesizes high-resolution satellite images from text prompts, improving detail and accuracy over existing methods in remote sensing imagery generation.

Contribution

Introduces a novel two-stage diffusion approach combining low-resolution generation and super-resolution refinement for satellite image synthesis from text.

Findings

Outperforms existing models in accuracy and resolution

Produces geographically detailed satellite images

Enhances spatial clarity in generated images

Abstract

The generation and enhancement of satellite imagery are critical in remote sensing, requiring high-quality, detailed images for accurate analysis. This research introduces a two-stage diffusion model methodology for synthesizing high-resolution satellite images from textual prompts. The pipeline comprises a Low-Resolution Diffusion Model (LRDM) that generates initial images based on text inputs and a Super-Resolution Diffusion Model (SRDM) that refines these images into high-resolution outputs. The LRDM merges text and image embeddings within a shared latent space, capturing essential scene content and structure. The SRDM then enhances these images, focusing on spatial features and visual clarity. Experiments conducted using the Remote Sensing Image Captioning Dataset (RSICD) demonstrate that our method outperforms existing models, producing satellite images with accurate geographical details and improved spatial resolution.