Learning Enriched Features for Real Image Restoration and Enhancement
Motivation
- Full-resolution representation is spatially precise but contextually less robust.
- Progressively low-resolution representation semantically reliable but spatially less accurate.
这篇文章提出在保持spatially precise同时也提取出具有 strong contextual information的特征。
提出的module 主要包含了如何提取多尺度的信息，多尺度信息的交换，以及空间通道注意力机制。
Existing CNN-based methods typically operate either on full-resolution or on progressively low-resolution representations
Method:
