The paper introduces a novel underwater image enhancement network called PDCFNet, which is based on pixel difference convolution (PDC) and cross-level feature fusion. The key highlights are:
PDCFNet includes a Detail Enhancement Module (DEM) that utilizes parallel PDCs to capture high-frequency features, leading to better detail and texture enhancement in underwater images.
The Feature Fusion Module (FFM) in PDCFNet performs operations like concatenation and multiplication on features from different levels, ensuring sufficient interaction and enhancement among diverse features.
Comprehensive experiments on three public datasets demonstrate that PDCFNet outperforms state-of-the-art underwater image enhancement methods in both quantitative and qualitative evaluations. It achieves the best performance in terms of metrics like MSE, PSNR, and SSIM.
PDCFNet is effective in addressing various underwater degradation scenarios, such as color deviation, blurriness, scattering, and darkness. It can restore colors effectively and improve the overall visibility of underwater images.
Ablation studies confirm the importance of the proposed PDC and the multi-loss function in achieving the superior performance of PDCFNet.
A otro idioma
del contenido fuente
arxiv.org
Ideas clave extraídas de
by Song Zhang, ... a las arxiv.org 10-01-2024
https://arxiv.org/pdf/2409.19269.pdfConsultas más profundas