Skip to main content

Table 1 Brief description of the algorithms.

From: Underwater Image Processing: State of the Art of Restoration and Image Enhancement Methods

Algorithm Model's characteristics and assumptions Experiments and data set Image quality evaluation
Jaffe [2] 1990 Computer modeling. Image as linear superposition of direct, forward and scattered components. Essentially for artificial lit scenes. Simulation and utility of different imaging and lighting configurations are evaluated. Visual inspection.
Image restoration methods
Hou et al. [9, 10] 2007 Measurement of PSF of water and automated restoration scheme. Natural and artificial lighting. Blurring caused by strong scattering due to water and floating particles. Two water types (clear and turbid), morning and afternoon. Target between 3.7 and 7.1 m. Visual inspection. Image quality metric: Weighted Gray Scale Angles (WGSA).
Trucco and Olmos [13] 2006 Self tuning restoration filter. Uniform illumination Only forward scatter is considered. Limited backscatter. Ocean images in shallow water, direct sunlight illumination. Some images with high backscatter. Visual inspection. Quantitative tests on frames from real mission videos. Improvement to classification tasks for subsea operations (detecting man-made objects on the seafloor).
Liu et al. [16] 2001 Measurement of PSF of water and image restoration. Standard and parametric Wiener filter deconvolution. Measurements on controlled environment. Set up: light source, slit images at 1–3 m in water tank. Restoration of images taken in turbid water. Visual inspection.
Schechner and Karpel [17] 2005 Polarization associated with the prime visibility disturbance to be deleted (backscatter). Natural lighting. Polarizer used to analyze the scene. Experiments in the sea (scene 26 m deep). Visual inspection. Quantitative estimate for the visibility improvement. Estimation of the distance map of the scene.
Treibitz and Schechner [19] 2009 Polarization-based method for visibility enhancement and distance estimation in scattering media. Artificial illumination. Experiments in real underwater scenes: Mediterranean sea, Red Sea and lake of Galilee. Visual inspection. Quantitative estimate for the visibility improvement.
Image enhancement and color correction methods
Bazeille et al. [20] 2006 Automatic pre-processing. Natural and artificial illumination. Deep marine habitats. Scenes with man-made objects in the sea floor. Visual inspection. Quantitative index: closeness of histogram to exponential distribution and tests for object recognition in the sea floor.
Chambah et al. [23] 2004 Underwater color constancy. Artificial lighting. Images taken in aquariums. Tests on fish segmentation and fish recognition. Visual inspection.
Iqbal et al. [25] 2007 Enhancement based on slide stretching. Natural and artificial illumination. Marine habitats. Visual inspection and histogram analysis.
Arnold-Bos et al. [22, 26] 2005 Automatic free denoising. Backscatter is considered as the first noise. Adaptive smoothing filter. Natural and artificial lighting. Marine habitats with unknown turbidity characteristics. Visual inspection. Quantitative criteria based on closeness of histogram to exponential distribution.
Torrez-Mendez and Dudek [27] 2005 Color recovery using an energy minimization formulation. Natural and artificial lighting. Training data set: marine habitats (ground truth is known) and frames from videos in the deep ocean (no ground truth available). Residual error is computed between ground truth and corrected images.
Ahlen et al. [28] 2007 Hyperspectral data for color correction. Natural illumination. Test image: colored plate at 6 m depth in the sea. Coral reefs and marine habitats. Visual inspection.
Petit et al. [29] 2009 Enhancement method: color space contraction using quaternions. Natural and artificial lighting Marine habitats at both shallow and deep waters. Visual inspection.
Garcia et al. [30] 2002 Compensating for lighting problems: non uniform illumination. Shallow waters on a sunny day. Shallow waters at sun down (simulating deep ocean). Visual inspection.
Arredondo and Lebart [39] 2005 Video processing algorithms. Simulations of perturbations. Natural and artificial lighting Test images are degraded with simulated perturbations. Simulations in shallow (1–7 m) and deep waters. Visual inspection. Quantitative evaluation: mean angular error is measured in motion estimation for different methods as a function of Gaussian noise.