Approximate Fast Foreground Colour Estimation

Official repository for the paper Approximate Fast Foreground Colour Estimation. ICIP 2021.

Naively compositing without an estimate of foreground colour leads to bleed-through of the original background.

In this work, we propose a foreground estimation technique. Our method can be implemented 11 lines of Python code and yields comparable results to the full approach [1], while also being faster.

• Watch project video

Notebooks

We provide notebooks to reproduce the results, and figures found in the paper.

Requirements

• numpy # For mathematical operations.
• opencv-python # Efficient blur implementation.
• torch >= 1.4 # For GPU computation.

Citation

@INPROCEEDINGS{ApproxForegroundForte2021,
  author={Marco Forte},
  booktitle={2021 IEEE International Conference on Image Processing (ICIP)}, 
  title={Approximate Fast Foreground Colour Estimation}, 
  year={2021},
  doi={10.1109/ICIP42928.2021.9506164}}

References

(1) Fast Multi-Level Foreground Estimation