The Color Improvement of Underwater Images Based on Light Source and Detector

Xiangqian Quan, Yucong Wei, Bo Li, Kaibin Liu, Chen Li, Bing Zhang, Jingchuan Yang2022 · DOI

TL;DR

This paper tackles the persistent problem of color distortion in underwater images, which hinders crucial marine applications. Instead of relying solely on post-processing, the authors propose a hardware-based approach by optimizing the light source and camera detector's spectral characteristics, demonstrating significant color improvement with retained image detail and reduced processing time.

What Problem Does This Paper Address?

Underwater environments present a unique challenge for optical imaging. Water acts as a natural filter, selectively attenuating different wavelengths of light. This attenuation, coupled with scattering, leads to significant color shifts and loss of detail in captured images. These distortions compromise the accuracy and efficiency of various underwater applications, including marine geology surveys, underwater mining, fishery, and marine archaeology. Existing solutions often involve complex image processing algorithms that can be computationally expensive and may lead to loss of image information. This paper aims to address this by proposing a more fundamental, hardware-centric solution that tackles the root cause of color deviation.

Methods

The core of this research lies in a hardware-based approach to color correction. The authors propose adjusting two key components of the imaging system:

  • Light Source Spectral Component: The paper investigates how different light sources, characterized by their spectral output, affect the color rendition of underwater images. They hypothesize that a light source with a specific spectral distribution can mitigate color distortion.
  • Detector Spectral Response: Similarly, they propose modifying the spectral response of the camera's detector. This involves designing or selecting detectors that are more sensitive to the wavelengths that are less attenuated in water or that can better compensate for the attenuation.

To validate their approach, they built an experimental setup that allowed them to systematically vary light sources and cameras while capturing images of a standard color chart underwater. They then analyzed the color deviation using the CIELAB color space, focusing on the 'a' and 'b' parameters, which represent color information.

Datasets

The paper does not explicitly mention training datasets. Instead, it focuses on experimental data acquisition. They captured 144 underwater images and extracted 1008 color blocks from a standard 24-color board under various conditions (different distances, lamps, and cameras). This dataset was used for quantitative analysis of color deviation.

Key Insights and Ideas

  • Hardware-Centric Solution: The most significant insight is the shift from purely software-based color correction to a hardware-based approach. By optimizing the light source and detector, they aim to capture more accurate color information from the outset, rather than trying to "fix" it later.
  • Warm Light Advantage: The experiments revealed that a warm-light LED exhibited the smallest color deviation among the tested lamps. This suggests that a specific spectral composition of the illumination can significantly improve color reproduction.
  • Novel 3xCMOS RGB Camera: The development of a novel 3xCMOS RGB camera, which uses a prism to split light into R, G, and B channels and then combines data from three separate detectors, proved highly effective. This camera design, with modifications to the prism's spectral transmittance (specifically enhancing the red channel), demonstrated the smallest color deviation among the tested cameras.
  • Efficiency Gains: The hardware-based approach offers practical advantages: retaining more image information and consuming less processing time, which are crucial for real-time underwater applications.

Key Results

The experimental results strongly support the proposed hardware-based approach:

  • Light Source: The warm-light LED resulted in the smallest color deviation, with a calculated value of $\sqrt{\Delta a^{*2} + \Delta b^{*2}} = 26.58$. This was significantly better than other tested lamps.
  • Detector: The custom-designed 3xCMOS RGB camera achieved the lowest color deviation, with a value of $\sqrt{\Delta a^{*2} + \Delta b^{*2}} = 25.25$. This outperformed standard consumer cameras.
  • General Trend: Color deviation was observed to increase with imaging distance, highlighting the importance of addressing attenuation.

Limitations and Future Work

  • Limited Lamp and Camera Variety: While the study tested several lamps and cameras, the scope was limited. Future work could explore a wider range of lighting technologies and camera sensor designs.
  • Specific Water Conditions: The experiments were conducted with tap water. The attenuation coefficient of water can vary significantly depending on salinity, turbidity, and dissolved substances. Future research could investigate the effectiveness of the proposed method across different water types.
  • Real-World Underwater Testing: While the setup simulated underwater conditions, extensive testing in diverse, real-world underwater environments would be beneficial to validate the robustness of the approach.
  • Integration with Image Processing: The authors suggest that their hardware-based method could be combined with image processing techniques for even greater color correction, which could be a promising avenue for future research.

Summary updated

BibTeX
@misc{quan2022color,
  author = {Xiangqian Quan and Yucong Wei and Bo Li and Kaibin Liu and Chen Li and Bing Zhang and Jingchuan Yang},
  title = {{The Color Improvement of Underwater Images Based on Light Source and Detector}},
  year = {2022},
  doi = {10.3390/s22020692},
}
Figure