Scaling Up Image Restoration: Achieving Photo-Realistic Results with Large-Scale Generative Models and Multimodal Guidance
The core message of this paper is that by scaling up the generative model and training data, and incorporating multimodal textual guidance, the authors have developed a powerful image restoration method called SUPIR that can achieve exceptional photo-realistic results, especially on complex real-world scenarios.
Dynamic Pre-training for Efficient and Scalable All-in-One Image Restoration
The core message of this paper is to introduce a novel weight-sharing mechanism within a Dynamic Network (DyNet) architecture that enables efficient and scalable all-in-one image restoration, significantly improving computational efficiency while boosting performance.
Towards Blind Image Restoration with Generative Diffusion Prior
DiffBIR decouples the blind image restoration problem into two stages: 1) degradation removal and 2) information regeneration, and leverages the generative ability of latent diffusion models to achieve state-of-the-art performance for blind super-resolution, blind face restoration, and blind image denoising tasks.
Deep Equilibrium Diffusion Models for Efficient and Controllable Image Restoration
Diffusion model-based image restoration can be formulated as a deep equilibrium fixed point system, enabling parallel sampling and efficient gradient computation for improved performance and controllability.
Under-Display Camera Image Restoration with Segmentation Guided Sparse Transformer
Transformer-based Sparse Attention improves UDC image restoration by filtering noise and focusing on relevant features.
Always Clear Days: Degradation Type and Severity Aware All-In-One Adverse Weather Removal
提案されたUtilityIRモデルは、天候の劣化タイプと深刻さを認識し、単一画像に対する盲目的な全天候除去を可能にします。
IRConStyle: Image Restoration Framework Using Contrastive Learning and Style Transfer
Contrastive learning paradigm applied to image restoration can significantly enhance performance by integrating style transfer and ConStyle module.
Efficient Diffusion Model for Image Restoration by Residual Shifting
The author proposes an efficient diffusion model tailored for image restoration, reducing the number of diffusion steps while maintaining performance. The model balances fidelity and perceptual quality through hyperparameters tuning.
Degradation Type and Severity Aware All-In-One Adverse Weather Removal
The author proposes a method that is type and severity aware for blind all-in-one weather removal, utilizing Contrastive Loss (CL) and Marginal Quality Ranking Loss (MQRL) to guide the model in extracting representative weather information effectively.
© 2024 by Linnk AI