[Paper Review] Effective Backscatter Approximation for Photometry in Murky Water

Chourmouzios Tsiotsios, Maria E. Angelopoulou+2 more|arXiv (Cornell University)|Apr 29, 2016

Water Quality Monitoring Technologies1 references3 citations

TL;DR

This paper proposes a novel backscatter approximation method for photometric stereo in murky water, leveraging the depth-saturation and smooth angular variation of backscatter to estimate it directly from captured images. By removing the complex backscatter component, the method enables accurate dense normal and albedo estimation, even under strong scattering, outperforming prior approaches in both simulations and real-world port water experiments.

ABSTRACT

Shading-based approaches like Photometric Stereo assume that the image formation model can be effectively optimized for the scene normals. However, in murky water this is a very challenging problem. The light from artificial sources is not only reflected by the scene but it is also scattered by the medium particles, yielding the backscatter component. Backscatter corresponds to a complex term with several unknown variables, and makes the problem of normal estimation hard. In this work, we show that instead of trying to optimize the complex backscatter model or use previous unrealistic simplifications, we can approximate the per-pixel backscatter signal directly from the captured images. Our method is based on the observation that backscatter is saturated beyond a certain distance, i.e. it becomes scene-depth independent, and finally corresponds to a smoothly varying signal which depends strongly on the light position with respect to each pixel. Our backscatter approximation method facilitates imaging and scene reconstruction in murky water when the illumination is artificial as in Photometric Stereo. Specifically, we show that it allows accurate scene normal estimation and offers potentials like single image restoration. We evaluate our approach using numerical simulations and real experiments within both the controlled environment of a big water-tank and real murky port-waters.

Motivation & Objective

• Address the challenge of backscatter degradation in underwater photometric stereo, which distorts image formation and hinders normal and albedo estimation.
• Overcome limitations of prior methods that either neglect backscatter or assume unrealistic lighting configurations (e.g., distant sources).
• Enable accurate scene reconstruction in murky water using artificial illumination, particularly for textureless or low-feature surfaces.
• Demonstrate the feasibility of single-image restoration and robust photometric stereo under strong scattering conditions.
• Provide a practical, calibration-free method for backscatter estimation directly from sensor data, suitable for real-time robotic underwater systems.

Proposed method

• Observe that backscatter saturates with scene depth beyond a short distance, making it independent of depth and dependent only on light-source-to-pixel geometry.
• Model the backscatter signal as a smooth function across the image plane, varying with the angular position of each pixel relative to the light source.
• Approximate the per-pixel backscatter using low-order polynomial regression (e.g., quadratic) trained on a few measured backscatter values from the image.
• Alternatively, estimate backscatter by capturing an image with the camera pointing at infinity (uniform illumination), which directly measures the backscatter level.
• Subtract the estimated backscatter from the measured brightness to isolate the scene-reflected component, enabling standard photometric stereo optimization.
• Apply the method in both calibrated and uncalibrated settings, supporting both multi-view photometric stereo and single-image restoration.

Experimental results

Research questions

• RQ1 Can backscatter in murky water be effectively approximated without modeling its complex physical dependencies on medium properties and depth?
• RQ2 How does the angular dependence of backscatter on pixel position relative to the light source affect image formation and reconstruction quality?
• RQ3 Can a smooth, low-dimensional approximation of backscatter enable accurate photometric stereo reconstruction in high-scattering environments?
• RQ4 Does the proposed method outperform existing approaches that neglect backscatter or assume unrealistic lighting configurations?
• RQ5 Can the same backscatter estimation framework be extended to single-image restoration without prior knowledge of medium or lighting parameters?

Key findings

• Backscatter becomes saturated with depth beyond a short distance, rendering it independent of scene depth and simplifying its modeling.
• The backscatter signal varies smoothly across the image plane based on the angular relationship between each pixel and the light source, enabling regression-based approximation.
• The proposed method achieves photometric stereo reconstruction quality comparable to clean-water conditions, even under strong scattering in real port water.
• In controlled water-tank experiments, the method outperforms prior approaches that neglect backscatter or assume distant lighting, especially at high scattering levels.
• Single-image restoration using the backscatter approximation successfully recovers contrast and albedo, matching or exceeding results from polarizer-based methods.
• The method is robust and practical for real-world ROV deployment, as demonstrated by successful reconstruction of detailed surface textures (e.g., shells) in murky port water.