Abstract:This paper introduces a new benchmarking dataset for marine snow removal of underwater images. Marine snow is one of the main degradation sources of underwater images that are caused by small particles, e.g., organic matter and sand, between the underwater scene and photosensors. We mathematically model two typical types of marine snow from the observations of real underwater images. The modeled artifacts are synthesized with underwater images to construct large-scale pairs of ground truth and degraded images to calculate objective qualities for marine snow removal and to train a deep neural network. We propose two marine snow removal tasks using the dataset and show the first benchmarking results of marine snow removal. The Marine Snow Removal Benchmarking Dataset is publicly available online.

What problem does this paper attempt to address?

The paper primarily addresses a specific issue in underwater image processing—Marine Snow removal. Marine Snow refers to the image degradation phenomenon caused by small particulate matter (such as organic matter, sand particles, etc.) in underwater environments. These particles reflect light and create interfering spots or streaks in the images. The paper tackles this problem through the following aspects: 1. **Dataset Construction**: To evaluate and train algorithms for Marine Snow removal, the authors created a new benchmark dataset called the "Marine Snow Removal Benchmarking Dataset" (MSRB dataset). This dataset contains a large number of synthetic underwater image pairs, each pair consisting of an original clear image and an image with simulated Marine Snow degradation effects. 2. **Mathematical Modeling**: Based on actual observations, the authors proposed two typical Marine Snow models: Highland Type and Volcanic Crater Type. These models are used to simulate the pixel value distribution characteristics of Marine Snow in digital images and generate synthetic Marine Snow accordingly. 3. **Task Definition**: The paper defines two specific tasks for Marine Snow removal. The first task (Task 1) focuses on removing small-sized Marine Snow; the second task (Task 2) covers the removal of Marine Snow of various sizes, from small to large. Task 2 is more challenging. 4. **Algorithm Evaluation**: The authors evaluated the performance of several different methods using the newly created dataset, including traditional Median Filters, improved Median Filters, and deep learning-based methods (such as U-Net). Experimental results show that deep learning-based methods perform the best in removing Marine Snow. 5. **Real-World Testing**: In addition to testing on the synthetic dataset, the authors also demonstrated the results of Marine Snow removal on real underwater images, verifying the effectiveness of their methods in practical applications. In summary, this paper effectively addresses the issue of Marine Snow removal in underwater images by constructing a high-quality dataset, proposing reasonable mathematical models for Marine Snow, and employing advanced deep learning techniques.

Marine Snow Removal Benchmarking Dataset

A deep learning approach for marine snow synthesis and removal

A Dataset for Detection and Segmentation of Underwater Marine Debris in Shallow Waters

Coastal Fisheries Resource Monitoring Through A Deep Learning-Based Underwater Video Analysis

A Multi-Stage Progressive Network with Feature Transmission and Fusion for Marine Snow Removal

Underwater image sharpening and color correction via dataset based on revised underwater image formation model

An Underwater Image Enhancement Benchmark Dataset and Beyond

OceanBench: The Sea Surface Height Edition

Physics-Inspired Synthesized Underwater Image Dataset

Machine Learning for Shipwreck Segmentation from Side Scan Sonar Imagery: Dataset and Benchmark

Marine vessel detection dataset and benchmark for unmanned surface vehicles

Ship Detection for SAR Imagery Based on Deep Learning: A Benchmark

A Benchmark dataset for both underwater image enhancement and underwater object detection

Dataset and Benchmark for Ship Detection in Complex Optical Remote Sensing Image

Marine Video Kit: A New Marine Video Dataset for Content-based Analysis and Retrieval

A Holistic Marine Video Dataset

Optimized Custom Dataset for Efficient Detection of Underwater Trash

SafeSea: Synthetic Data Generation for Adverse & Low Probability Maritime Conditions

Semantic Segmentation of Underwater Imagery: Dataset and Benchmark

Image Blending-Based Noise Synthesis and Attention-Guided Network for Single Image Marine Snow Denoising

An Efficient Deep-Sea Debris Detection Method Using Deep Neural Networks