SPIRE: Semantic Prompt-Driven Image Restoration

TL;DR

SPIRE introduces a novel language-driven image restoration framework that uses natural language prompts for semantic control and fine-level instruction, achieving superior results across various restoration tasks.

Contribution

It is the first framework to support language-based quantitative control of restoration strength and enhances ControlNet with a new fusion mechanism for better fidelity.

Findings

Outperforms state-of-the-art restoration methods

Supports fine-level language instructions for control

Demonstrates flexible, text-based image restoration

Abstract

Text-driven diffusion models have become increasingly popular for various image editing tasks, including inpainting, stylization, and object replacement. However, it still remains an open research problem to adopt this language-vision paradigm for more fine-level image processing tasks, such as denoising, super-resolution, deblurring, and compression artifact removal. In this paper, we develop SPIRE, a Semantic and restoration Prompt-driven Image Restoration framework that leverages natural language as a user-friendly interface to control the image restoration process. We consider the capacity of prompt information in two dimensions. First, we use content-related prompts to enhance the semantic alignment, effectively alleviating identity ambiguity in the restoration outcomes. Second, our approach is the first framework that supports fine-level instruction through language-based quantitative specification of the restoration strength, without the need for explicit task-specific design. In addition, we introduce a novel fusion mechanism that augments the existing ControlNet architecture by learning to rescale the generative prior, thereby achieving better restoration fidelity. Our extensive experiments demonstrate the superior restoration performance of SPIRE compared to the state of the arts, alongside offering the flexibility of text-based control over the restoration effects.