Research map / Architectures and efficiency / CNN design
Learning Sensor Multiplexing Design through Back-propagation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 traditional demosaicking with the learned reconstruction network on the Bayer pattern | Bayer pattern · Median PSNR, noise STD=0.01 | PSNR (dB): 32.44 -> 43.72 | Helped |
| Replace traditional reconstruction with the network on the Bayer pattern at low noise | Bayer pattern · Median PSNR, noise STD=0.0025 | PSNR (dB): 42.69 -> 47.55 | Helped |
| Replace the CFZ method's own reconstruction with the network at low noise | CFZ pattern · Median PSNR, noise STD=0.0025 | PSNR (dB): 48.84 -> 49.08 | No clear effect |
| Replace CFZ's own reconstruction with the network at moderate noise | CFZ pattern · Median PSNR, noise STD=0.01 | PSNR (dB): 39.55 -> 44.64 | Helped |
| Use the learned sensor pattern instead of Bayer, with the same network | Reconstruction network · Median PSNR, noise STD=0.01 | PSNR (dB): 43.72 -> 44.94 | Helped |
| Use the learned sensor pattern instead of Bayer at high noise | Reconstruction network · Median PSNR, noise STD=0.04 | PSNR (dB): 39.98 -> 41.02 | Helped |
| Use the learned pattern instead of CFZ pattern at moderate noise | Reconstruction network · Median PSNR, noise STD=0.01 | PSNR (dB): 44.64 -> 44.94 | Helped |
| Use the learned pattern instead of CFZ at zero noise (median) | Reconstruction network · Median PSNR, noise STD=0 | PSNR (dB): 52.17 -> 52.12 | No clear effect |
| Use the learned pattern instead of CFZ for the 75th percentile at the highest noise | Reconstruction network · 75th percentile PSNR, noise STD=0.04 | PSNR (dB): 45.11 -> 44.98 | Hurt |
| Train the sensor pattern at only STD=0.01 and test across other noise levels | Learned pattern · PSNR across noise STD 0 to 0.04 | PSNR (dB) | Helped |
| Train longer: sensor layer shifts from color sampling to panchromatic pixels, with decreasing mean entropy of selection map | Sensor layer · Pattern training dynamics | selection-map entropy | Helped |
Closest papers by meaning
- Leveraging Multispectral Sensors for Color Correction in Mobile Cameras (CVPR 2026 · similarity 0.84)
- Model-Based Image Signal Processors via Learnable Dictionaries (AAAI 2022 · similarity 0.84)
- Self-Supervised Learning for Color Spike Camera Reconstruction (CVPR 2025 · similarity 0.83)
- GCC: Generative Color Constancy via Diffusing a Color Checker (CVPR 2025 · similarity 0.82)
- Efficient Neural Network Encoding for 3D Color Lookup Tables (AAAI 2025 · similarity 0.81)
- How Hard Is It for Message-Passing GNNs to Simulate One Weisfeiler-Lehman Color-Refinement Step? (NeurIPS 2026 · similarity 0.81)
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.