摘要 Abstract 

在本次演講中，我們將介紹影像還原與增強的最新進展，聚焦於三個主要任務：影像去噪、去雨與增強。針對去噪，我們提出SUNet，為首個結合Swin-Transformer與U-Net的架構；並提出了SRMNet，一種能同時處理合成雜訊與真實雜訊的盲去噪網路。在去雨方面，我們提出CMFNet，模型靈感來自視網膜神經節細胞，可修復被雨滴破壞的影像。此外，我們提出HWMNet，一種針對低光影像增強所設計的改良式階層架構。

In this talk, we will present recent advances in image restoration and enhancement, focusing on three key tasks: image denoising, deraining and enhancement. For denoising, we propose SUNet, the first architecture combining Swin-Transformer and U-Net, and SRMNet, a blind denoising network capable of handling both synthetic noise and real-world noise. For deraining, we introduce CMFNet, a model inspired by retinal ganglion cells to restore images degraded by raindrops. Additionally, we present HWMNet, an improved hierarchical architecture designed for low-light image enhancement.