Research map / Image generation / Image restoration
RefocusGAN: Scene Refocusing using a Single Image
Parikshit Sakurikar, Ishit Mehta, Vineeth N. Balasubramanian, P. J. Narayanan
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 SML focus measure channel to the deblurring input (ablation) | RefocusGAN deblurring cGAN · Deblurring (light-field test split) | PSNR: 34.88 -> 35.02 | Improved |
| add SML focus measure channel to the deblurring input (ablation) | RefocusGAN deblurring cGAN · Deblurring (light-field test split) | SSIM: 0.937 -> 0.938 | Improved |
| add the deblurred in-focus radiance image as input to the refocusing network (two-stage) (ablation) | RefocusGAN refocusing cGAN, delta=+8 · Refocusing (delta=+8) | PSNR: 38.73 -> 44.225 | Improved |
| add the deblurred in-focus radiance image as input to the refocusing network (two-stage) (ablation) | RefocusGAN refocusing cGAN, delta=+8 · Refocusing (delta=+8) | SSIM: 0.97 -> 0.992 | Improved |
| add the deblurred in-focus radiance image as input to the refocusing network (two-stage) (ablation) | RefocusGAN refocusing cGAN, four deltas · Refocusing (delta in {-9,-5,+5,+9}) | PSNR: 38.4 -> 43.4 | Improved |
| add the deblurred in-focus radiance image as input to the refocusing network (two-stage) (ablation) | RefocusGAN refocusing cGAN, four deltas · Refocusing (delta in {-9,-5,+5,+9}) | SSIM: 0.956 -> 0.988 | Improved |
| increase the number of refocus control parameters from 4 to 19 (ablation) | RefocusGAN refocusing cGAN · Refocusing (two-stage) | PSNR: 43.4 -> 40.42 | Hurt |
| increase the number of refocus control parameters from 4 to 19 (ablation) | RefocusGAN refocusing cGAN · Refocusing (two-stage) | SSIM: 0.988 -> 0.975 | Hurt |
| train on a single refocus parameter (+8) instead of four (ablation) | RefocusGAN refocusing cGAN · Refocusing (two-stage) | PSNR: 43.4 -> 44.225 | Improved |
| test with network-computed radiance instead of ground-truth in-focus images (model trained on ground truth) (ablation) | RefocusGAN refocusing cGAN, four deltas · Refocusing (delta in {-9,-5,+5,+9}) | PSNR: 43.4 -> 38.63 | Hurt |
| use composite focus measure over the whole focal stack instead of single-image deblurring | in-focus image estimation · Deblurring (light-field test split) | PSNR: 35.02 -> 38.697 | Improved |
Closest papers by meaning
- Defocus Deblurring Using Dual-Pixel Data (ECCV 2020 · similarity 0.87)
- Quad-Pixel Image Defocus Deblurring: A New Benchmark and Model (CVPR 2025 · similarity 0.84)
- Exploiting Deblurring Networks for Radiance Fields (CVPR 2025 · similarity 0.84)
- Residual Diffusion Deblurring Model for Single Image Defocus Deblurring (AAAI 2025 · similarity 0.84)
- Dehaze-RetinexGAN: Real-World Image Dehazing via Retinex-based Generative Adversarial Network (AAAI 2025 · similarity 0.84)
- Multi-Focus Image Fusion via Explicit Defocus Blur Modelling (AAAI 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.
