Research map / 3D vision / Depth and stereo
MambaSIC: Mamba-based Stereo Image Compression with Bi-directional Multi-reference Entropy Model
Shiyu Qin, Xinjie Zhang, Zhening Liu, Jinpeng Wang, Bin Chen, Jiawei Li, Yifan Ren, Shu-Tao Xia, Jun Zhang
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 cross-view matrix alpha C to Stereo 2DSS (ablation) | MambaSIC · InStereo2K | BDBR PSNR: 3.86% bitrate increase when removed | Improved |
| add stereo gating connection (ablation) | MambaSIC Stereo VSSL · InStereo2K | BDBR PSNR: 6.98% bitrate increase when removed | Improved |
| use two-view stereo VSSB instead of single-view VSSB (ablation) | MambaSIC · InStereo2K | BDBR PSNR: 10.13% bitrate increase with single VSSB | Improved |
| add inter-view priors to the entropy model (ablation) | MambaSIC entropy model · InStereo2K | BDBR PSNR: 11.67% bitrate increase when removed | Improved |
| inject cross-view information through matrix C rather than B, Delta or output v (ablation) | Stereo 2DSS · InStereo2K | BD-PSNR: B -0.058dB, Delta -0.100dB, v -0.107dB | Improved |
| replace Stereo VSSB with BiSIC mutual attention (ablation) | MambaSIC · InStereo2K | BDBR PSNR and latency: BDBR +13.59%; latency 1.26 -> 2.64 | Hurt |
| use checkerboard bi-directional multi-reference entropy model instead of BiSIC spatial autoregressive entropy model (ablation) | MambaSIC · InStereo2K | coding latency (s): 75.19 -> 1.26 | Improved |
| use 2D convolutions instead of 3D convolutions with Mutual Attention backbone (ablation) | MambaSIC · InStereo2K | coding latency (s): 3.58 -> 1.26 | Improved |
| use BiSIC entropy model instead of ours (ablation) | MambaSIC · InStereo2K | BDBR PSNR: +4.05% bitrate | Hurt |
| MambaSIC replaces BiSIC as stereo codec | stereo image compression · InStereo2K | total coding latency (s): 78.6075 -> 1.2625 | Improved |
Closest papers by meaning
- 3D-LMVIC: Learning-based Multi-View Image Compression with 3D Gaussian Geometric Priors (ICML 2025 · similarity 0.87)
- CAMSIC: Content-aware Masked Image Modeling Transformer for Stereo Image Compression (AAAI 2025 · similarity 0.87)
- MambaIC: State Space Models for High-Performance Learned Image Compression (CVPR 2025 · similarity 0.86)
- MVSMamba: Multi-View Stereo with State Space Model (NeurIPS 2025 · similarity 0.86)
- Low-Latency Neural Stereo Streaming (CVPR 2024 · similarity 0.84)
- Spatial-Mamba: Effective Visual State Space Models via Structure-Aware State Fusion (ICLR 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.
