PHISWID
收藏资源简介:
PHISWID是由日本大阪大学创建的一个专为提升水下图像处理能力而设计的物理启发合成水下图像数据集。该数据集包含2264对图像,每对包括一个原始大气图像和一个通过颜色偏移和海洋雪效应降质的水下合成图像。数据集的创建过程涉及使用物理模型对海洋雪的光散射进行数学建模,并将其应用于大气RGB-D图像以增强数据集的真实性和适用性。PHISWID特别适用于监督学习环境下的深度神经网络训练,以及客观评估图像质量的基准分析,旨在解决水下图像增强中的挑战,特别是海洋雪效应的处理。
PHISWID is a physics-inspired synthetic underwater image dataset developed by Osaka University, Japan, specifically designed to enhance underwater image processing capabilities. This dataset comprises 2264 image pairs, each consisting of one raw atmospheric RGB-D image and one synthetic underwater image degraded via color shifting and marine snow effects. The construction of the dataset involves mathematically modeling the light scattering of marine snow based on physical principles, and applying this model to atmospheric RGB-D images to enhance the dataset's authenticity and practical applicability. PHISWID is particularly suitable for training deep neural networks in supervised learning scenarios, as well as conducting benchmark analyses for objective image quality assessment. It aims to address the challenges in underwater image enhancement, especially the handling of marine snow-induced degradation.




