Marine Snow Removal Benchmarking Dataset

Marine Snow Removal Benchmarking Dataset（MSRB数据集）是由日本大阪大学和东京农业技术大学合作创建的，专注于去除水下图像中的海洋雪粒子。该数据集包含5400对图像，每对包括一个原始水下图像和一个带有合成海洋雪粒子的图像。数据集的创建过程涉及从真实水下图像中观察并数学建模海洋雪粒子，然后将其合成到图像中。MSRB数据集主要用于训练深度神经网络，以解决水下图像增强中的海洋雪去除问题，特别是在极端图像恢复领域中。

The Marine Snow Removal Benchmarking Dataset (MSRB Dataset) was co-created by Osaka University and Tokyo University of Agriculture and Technology in Japan, focusing on the removal of marine snow particles from underwater images. This dataset contains 5400 image pairs, each including one raw underwater image and one image with synthetic marine snow particles. The creation of the MSRB Dataset entails observing marine snow particles from real-world underwater images, developing their mathematical models, and subsequently synthesizing these particles into the original images. The MSRB Dataset is primarily utilized for training deep neural networks to tackle the marine snow removal task in underwater image enhancement, especially within the domain of extreme image restoration.

搜集汇总

背景与挑战

背景概述

MSRB数据集是一个由日本高校合作创建的水下图像处理数据集，包含5400对图像，用于模拟和去除海洋雪粒子。它基于真实水下观察和数学建模合成海洋雪，主要应用于训练深度神经网络以增强水下图像质量，专注于极端图像恢复任务。