Research map / Image generation / Image restoration
Parallax-Tolerant Unsupervised Deep Image Stitching
Lang Nie, Chunyu Lin, Kang Liao, Shuaicheng Liu, Yao Zhao
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 |
|---|---|---|---|
| replace homography-only alignment (UDIS warp) with homography+TPS warp | UDIS++ warp · UDIS-D warp (Average) | PSNR: 21.17 -> 25.43 | Improved |
| use homography then TPS instead of homography twice (ablation) | UDIS++ warp · UDIS-D warp (Average) | PSNR: 24.46 -> 25.43 | Improved |
| use homography then TPS instead of TPS twice (ablation) | UDIS++ warp · UDIS-D warp (Average) | PSNR: 25.31 -> 25.43 | Improved |
| add inter-grid and intra-grid distortion constraints (ablation) | UDIS++ warp · UDIS-D warp (Average) | PSNR: 25.54 -> 25.43 | No clear effect |
| add inter-grid and intra-grid distortion constraints (ablation) | UDIS++ warp · UDIS-D warp (Average) | SSIM: 0.841 -> 0.838 | No clear effect |
| add smoothness constraints on difference map and stitched image in composition (ablation) | UDIS++ composition · qualitative composition | seam discontinuity | Improved |
| run iterative warp adaption at test time | UDIS++ warp · cross-dataset and cross-resolution | artifact reduction | Improved |
| replace reconstruction-based composition with seam-driven mask composition | UDIS composition · Railtrack composition time | elapsed time (s), CPU: 304.963 -> 22.778 | Improved |
Closest papers by meaning
- Warping Residual Based Image Stitching for Large Parallax (CVPR 2020 · similarity 0.93)
- Pixel-Wise Warping for Deep Image Stitching (AAAI 2023 · similarity 0.92)
- Leveraging Line-Point Consistence To Preserve Structures for Wide Parallax Image Stitching (CVPR 2021 · similarity 0.92)
- Depth-Supervised Fusion Network for Seamless-Free Image Stitching (NeurIPS 2025 · similarity 0.91)
- Reconstructing the Image Stitching Pipeline: Integrating Fusion and Rectangling into a Unified Inpainting Model (NeurIPS 2024 · similarity 0.86)
- Object-level Geometric Structure Preserving for Natural Image Stitching (AAAI 2025 · similarity 0.85)
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.
