Single Underwater Image Restoration Using Adaptive Attenuation-Curve Prior (AACP)

Experimental results and Matlab executable file for IEEE TCAS-I paper: Single Underwater Image Restoration Using Adaptive Attenuation-Curve Prior. link

We propose a novel underwater image restoration method based on a non-local prior, namely, adaptive attenuation-curve prior. This prior relies on the statistical distribution of pixel values. That is, all pixel values of a clear image can be partitioned into several hundred distinct clusters in RGB space (see Fig. 1 (b)), and the pixel values in each cluster will be distributed on a curve with a power function form after attenuated by water in varying degrees (see Fig. 1 (d)). Specifically, we can estimate the transmission for each pixel according to its distribution on the curves.

Link of experimental results and code

Experimental Results

Matlab Executable File(Due to the project limitation, only the executable file can be provided)

Example results

The visual comparison of four underwater images with different scenes. (a) The original image. The restored results, and the corresponding waterlight and transmission yielded by: (b) MIP [4], (c) UDCP [8], (d) blurriness-based method [9], (e) fixed-attenuation-curve method [5], (f) the proposed method (adaptive attenuation-curve).

Citation

If you find this research project is useful for your research, please cite:

@article{wang2018single,
  title={Single underwater image restoration using adaptive attenuation-curve prior},
  author={Wang, Yi and Liu, Hui and Chau, Lap-Pui},
  journal={IEEE Transactions on Circuits and Systems I: Regular Papers},
  volume={65},
  number={3},
  pages={992--1002},
  year={2018},
  publisher={IEEE}
}