Research map / Image generation / Image restoration
Deep Texture and Structure Aware Filtering Network for Image Smoothing
Kaiyue Lu, Shaodi You, Nick Barnes
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add structure guidance (HED) to the filtering network (ablation) | filtering network · Generated texture dataset test (guidance ablation) | MSE: 0.0316 -> 0.0215 | Improved |
| add texture guidance (TPN) to the filtering network (ablation) | filtering network · Generated texture dataset test (guidance ablation) | MSE: 0.0316 -> 0.0118 | Improved |
| add texture guidance on top of structure guidance (ablation) | filtering network · Generated texture dataset test (guidance ablation) | MSE: 0.0215 -> 0.0059 | Improved |
| add structure guidance on top of texture guidance (ablation) | filtering network · Generated texture dataset test (guidance ablation) | MSE: 0.0118 -> 0.0059 | Improved |
| fine-tune the whole network jointly after separate training (ablation) | TSAFN · Generated texture dataset test (guidance ablation) | PSNR: 24.78 -> 25.07 | Improved |
| add structure guidance to the filtering network (ablation) | filtering network · Generated texture dataset test (guidance ablation) | PSNR: 20.32 -> 21.71 | Improved |
| add texture guidance only (no structure guidance) (ablation) | filtering network · Generated texture dataset test (guidance ablation) | SSIM: 0.7934 -> 0.8201 | Mixed |
| replace hand-crafted filters with TSAFN | DGF (best hand-crafted) · Generated texture dataset test (hand-crafted filters) | MSE: 0.1247 -> 0.0051 | Improved |
| replace best retrained deep filter with TSAFN | CEILNet · Generated texture dataset test (deep models retrained) | MSE: 0.0156 -> 0.0051 | Improved |
| use learned TPN instead of hand-crafted separation for texture extraction | SGTD (best separation filter) · Generated texture dataset test (texture extraction) | MSE (texture extraction): 0.1315 -> 0.0196 | Improved |
| use TPN on a new texture set | SGTD · 100 new images (texture extraction) | MSE (texture extraction): 0.1608 -> 0.0212 | Improved |
Closest papers by meaning
- Deep Filter Banks for Texture Recognition, Description, and Segmentation (IJCV 2016 · similarity 0.86)
- Semantic Filtering (CVPR 2016 · similarity 0.86)
- Deep Structure-Revealed Network for Texture Recognition (CVPR 2020 · similarity 0.86)
- Edge-aware Image Smoothing with Relative Wavelet Domain Representation (ICLR 2025 · similarity 0.85)
- A Generic Deep Architecture for Single Image Reflection Removal and Image Smoothing (ICCV 2017 · similarity 0.85)
- ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
