BriefGPT.xyz
Nov, 2016
基于持久同调引导的流形简化，超越深度残差学习的图像恢复方法
Beyond Deep Residual Learning for Image Restoration: Persistent Homology-Guided Manifold Simplification
HTML
PDF
Woong Bae, Jaejoon Yoo, Jong Chul Ye
TL;DR
本文提出了一种新的特征空间深度残差学习算法，通过将输入和标签流形解析地映射到特征空间来简化模型，从而在图像去噪和超分辨率任务上超越了现有的残差学习算法，并在NTIRE比赛中获得第三名。
Abstract
The latest
deep learning
approaches perform better than the state-of-the-art signal processing approaches in various
image restoration
tasks. However, if an image contains many patterns and structures, the perfor
→