고해상도 이미지 복원을 위한 비선형 확산 프로세스 기반의 다중 스케일 방법: 국소 노이즈 제거 및 스펙트럼 다중 스케일 기저 함수
비선형 확산 프로세스를 활용하여 노이즈를 제거하고 이미지의 주요 특징을 보존하는 다중 스케일 방법을 제안한다. 국소 노이즈 제거와 스펙트럼 다중 스케일 기저 함수 구축을 통해 효율적이고 정확한 이미지 복원을 달성한다.
Enhancing Image Contrast and Tone through Mathematical Modeling of Dichotomy
A new mathematical model based on the power function can effectively enhance image contrast and tone by modeling image tone dichotomy, improving underexposed, overexposed, and mixed exposure images.
Enhancing Bracket Image Restoration and Reconstruction with Flow-Guided Alignment and Improved Feature Aggregation
The proposed IREANet framework effectively restores high-quality high dynamic range (HDR) images from noisy, blurred, and low dynamic range multi-exposure RAW inputs by incorporating a flow-guided feature alignment module and an enhanced feature aggregation module.
Enhancing Low-Light Images with Differentiable Spatial Entropy
The core message of this work is to introduce a novel statistic-based objective function, called spatial entropy loss, to improve the perceptual quality of diffusion-based image restoration models for low-light enhancement.
Joint Conditional Diffusion Model for Effective Blind Image Restoration under Mixed Weather Degradations
A novel Joint Conditional Diffusion Model (JCDM) is proposed to effectively restore images degraded by complex combinations of weather conditions, such as rain, haze, and snow, without the need for explicit degradation identification or separation.
© 2024 by Linnk AI