Underwater Single Image Color Restoration Using Haze-Lines and a New Quantitative Dataset (1811.01343v3)

Abstract: Underwater images suffer from color distortion and low contrast, because light is attenuated while it propagates through water. Attenuation under water varies with wavelength, unlike terrestrial images where attenuation is assumed to be spectrally uniform. The attenuation depends both on the water body and the 3D structure of the scene, making color restoration difficult. Unlike existing single underwater image enhancement techniques, our method takes into account multiple spectral profiles of different water types. By estimating just two additional global parameters: the attenuation ratios of the blue-red and blue-green color channels, the problem is reduced to single image dehazing, where all color channels have the same attenuation coefficients. Since the water type is unknown, we evaluate different parameters out of an existing library of water types. Each type leads to a different restored image and the best result is automatically chosen based on color distribution. We collected a dataset of images taken in different locations with varying water properties, showing color charts in the scenes. Moreover, to obtain ground truth, the 3D structure of the scene was calculated based on stereo imaging. This dataset enables a quantitative evaluation of restoration algorithms on natural images and shows the advantage of our method.

• The paper introduces an enhanced haze-lines model that compensates for wavelength-dependent light attenuation underwater.
• The authors integrate global attenuation ratios from Jerlov water types with edge detection to robustly restore color.
• The new stereo image dataset enables quantitative evaluation of color accuracy and depth estimation, outperforming previous methods.

Analyzing Underwater Single Image Color Restoration Using Haze-Lines and New Quantitative Dataset

The paper "Underwater Single Image Color Restoration Using Haze-Lines and a New Quantitative Dataset" presents a novel approach for restoring color and improving contrast in underwater imagery, a complex problem owing to the non-uniform attenuation of light across different wavelengths underwater. The model developed in this paper addresses the challenges posed by such conditions by expanding upon the haze-lines approach, which is successfully utilized in terrestrial dehazing applications.

Methodology

In underwater environments, light absorption and scattering cause significant color distortion and low contrast in images. The paper proposes an enhanced model acknowledging the spectral dependency of light attenuation, capitalizing on prior knowledge of various water types' spectral profiles. This is achieved by introducing two global parameters: the attenuation ratios between the blue-red and blue-green color channels. These parameters are estimated using a pre-existing library of water types defined by Jerlov, which classifies water bodies based on their optical properties.

The authors employ a physical model similar to that used for single image dehazing, where the problem is transformed into one with spectrally uniform attenuation. Additionally, a unique veiling-light estimation method is implemented using edge detection to isolate areas without any distinct objects. A comprehensive evaluation of potential water types allows the algorithm to automatically select the most appropriate restoration result, leveraging the gray-world assumption to ensure color accuracy.

Dataset Contribution

A significant aspect of this research is the introduction of a new quantitative dataset promoting rigorous evaluation of underwater image restoration techniques. The dataset comprises stereo images of underwater scenes containing color charts taken at various locations with differing water properties. This aspect of the dataset permits both quantitative evaluation of color restoration and validation of depth estimation through true distance calculations derived from stereo imaging.

Results and Implications

The paper highlights substantial improvements in color restoration and transmission map accuracy compared to existing methods, including those that do not account for wavelength-dependent attenuation. The proposed method achieves a high correlation between true distances and estimated transmissions, evidenced by strong numerical evaluations detailed in the paper.

The implications of this research extend into practical applications in underwater visual tasks such as object recognition, navigation, and marine biology studies—where accurate color rendition and depth estimation are crucial. The approach facilitates the enhancement of automatic segmentation and feature matching in multi-view scenarios, crucial for activities like underwater mapping and exploration.

Future Directions

This paper sets the stage for further exploration into adaptive models that adjust to temporal and spatial variability in water properties, which could not only refine restoration accuracy but also enhance real-time image processing capabilities. Moreover, the dataset opens avenues for machine learning applications by providing a benchmark for training and evaluation of learning-based image restoration algorithms.

The paper robustly addresses the challenge of single underwater image restoration and lays a foundational method that other researchers and practitioners in computer vision and underwater imaging fields can build upon. Future research could expand the scope of this model, integrating real-time adaptability and leveraging machine learning techniques to deal with dynamic underwater environments more effectively.