A Statistical Learning-Based Method for Color Correction of Underwater Images
Luz Abril Torres Méndez, Gregory Dudek · Research in Computing Science · 2005
This paper addresses the problem of color correction of underwater images using statistical priors. Underwater images present a challenge when trying to correct the blue-green monochrome shift to bring out the color visible under full spectrum illumination in a transparent medium. We propose a learning-based Markov Random Field (MRF) model based on training from examples. Training images are small patches of color depleted and color images. The most probable color assignment to each pixel in the given color depleted image is inferred by using a non-parametric sampling procedure. Experimental results on a variety of underwater scenes demonstrate the feasibility of our method