Underwater image restoration method based on turbulence model

Abstract

The invention relates to an underwater image restoration method based on a turbulence model. The method includes the steps of transforming an input color underwater image from RGB (Red Green Blue) space to YCbCr color space; smoothing gray components through a bilateral filtering method and representing low-frequency profile components and high-frequency detail components with YL and YH respectively after bilateral filtering processing; representing the gray components of the degenerated underwater image with YL (x, y), transforming YL (x, y) into a Fourier transform domain and representing the Fourier transform domain with YL (u, v); obtaining a non-degenerated underwater image F (u, v) through a full-blind iterative restoration algorithm; obtaining a degenerated image based on the turbulent model to obtain a restoration results; transforming the restoration result back to a spatial domain again; performing color correction; reconstructing a restoration image. By means of the underwater image restoration method based on the turbulence model, the contrast, the noise, and the color cast of the underwater image are significantly improved.

Description

Technical field [0001] The invention relates to image restoration technology in the field of computer vision, in particular to a technology for improving the visual quality of degraded images captured by underwater cameras. Background technique [0002] The ocean covers 71% of the earth's surface, and it contains huge mineral resources, biological resources and energy and energy. my country is a large ocean country with a coastline of 18,000 kilometers and a land area of ​​3 million square kilometers. The development of the ocean and the exploration of the laws of the ocean are of great significance and value to our national economy and military. [0003] Research on autonomous underwater vehicles (AUVs) has also attracted more and more attention from many countries and regions. AUVs can not only be used in the exploration and development of offshore resources (pipeline laying, seabed investigation, drilling support, seabed construction, offshore drilling platforms, etc.) E...

AI Technical Summary

Problems solved by technology

This method only considers the influence of forward scattering on image quality in underwater imaging, and ignores the backscattering noise, so this method is not suitable for underwater image restoration obtained in offshore areas of muddy water pools

Andreas et al. [2] Based on the underwater imaging model, a framework for comprehensive preprocessing of underwater images is proposed. The estimation of the backscattered light components uses the Gaussian function convolution method. One of the problems of this algorithm is that the estimation of scattered light is too time-consuming.

Schechner et al. [3] A model-based underwater image restoration algorithm is proposed, which uses a polarization filter to estimate and remove the scattered light of the medium, but it needs to take more than two images in the same scene as prior knowledge, which will limit the real-time performance of video observation

Domestically, Liu et al. [4] The dark channel theory is used to approximate the depth information of the underwater environment, and then the restoration process is performed according to the underwater imaging model. The limitation of the algorithm is not suitable for the situation when the color of the scene object and the background light are similar.

Images