GPU-Accelerated Iterative Bilateral Solver

This repository provides a standalone implementation of the GPU-accelerated bilateral solver introduced in our paper (Modular Neural Image Signal Processing, Afifi et al., 2025 — supplemental Sec. A).

Our solver is built on the original Fast Bilateral Solver (FBS), but is redesigned entirely using dense GPU-friendly tensor operations, achieving up to 15x (~7x on average) faster performance on GPU^* compared to the CPU-based FBS.

^* Measured on an Intel Core i7-14700K CPU and an NVIDIA GeForce RTX 4080 SUPER (16 GB VRAM).

---

🛠️ Requirements

• Python 3.8+
• PyTorch (GPU build recommended)
• OpenCV
• NumPy

---

▶️ Running the Demo

python demo.py \
    --input-image-path path/to/input.png \
    --reference-image-path path/to/guidance.png

The solver will refine the input using the guidance image, save the output as: input_output.png

---

📄 Citation

If you use this solver, please cite:

@inproceedings{afifi2026modular,
  title={Modular Neural Image Signal Processing},
  author={Afifi, Mahmoud and Wang, Zhongling and Zhang, Ran and Brown, Michael S},
  booktitle = {SIGGRAPH Asia 2026 Conference Papers},
  year={2026}
}