PA-ColorNet: progressive attention network based on RGB and HSV color spaces to improve the visual quality of underwater images

Mengjun Miao,Shanqin Wang

DOI: https://doi.org/10.1007/s11760-023-02562-7

2023-04-14

Abstract:Due to the complexity of underwater scenes, underwater imaging is affected by the absorption and scattering of light through the water medium, resulting in degradation such as color casts, low contrast, and blurring. Many existing deep learning-based methods often fail to consider the impact of image degradation on different color channels, resulting in limited performance. In this work, we propose a progressive attention network based on RGB and HSV color spaces to improve the visual quality of underwater images. Specifically, we first transform the raw image into HSV color space and form RGB and HSV branches, respectively. We separate the color channels in each branch and employ the designed FFT-based aggregated residual dense module to focus on spatial and frequency domain features. Second, we feed the learned features of each channel on the two branches into the triple-color channel-wise attention module separately to balance the color distribution of different channels. Finally, we combine the results from two branches to obtain a high-quality image. Extensive qualitative and quantitative experimental results show that our model outperforms the compared methods. To evaluate the effectiveness of the components of our model, we conduct a series of ablation experiments.

engineering, electrical & electronic,imaging science & photographic technology