Underwater Image Enhancement Based On Structure-texture Decomposition

Underwater imaging technology is a new marine environment monitoring technology.Clear underwater image is very important for monitoring marine environmental,underwater survey,military reconnaissance and other aspects.However,the complex marine environment and a large number of suspended particles bring great difficulties and challenges to high clear imaging.Since the underwater imaging model is similar to the atmospheric imaging model,the image dehazing technology is applied to underwater image enhancement firstly,but it usually lead to oversaturation in some areas.Therefore,various models for underwater image enhancement have been proposed,however these methods still can not achieve satisfactory results in detail preservation,color correction,noise removal and so on.Aiming at these problem,this paper proposes an underwater image enhancement method based on structure-texture layer decomposition,in which the main achievements and innovations as follows:Ⅰ.Proposing the subregion background light estimation method which utilities respectively the threshold segmentation to estimate the highlight area and other areas of the image,so as to avoid the influence of artificial light source and over-bright area on the estimation results.Experiments show that the proposed algorithm can estimate the background light more accurately,thus improving the enhancement effect in the dark region.Ⅱ.Proposing an underwater image enhancement method based on structure texture decomposition.The total variation-norm minimization(TV-L1)algorithm is proposed to decompose the image into high-frequency texture layer and low-frequency structure layer to preserve the enhanced details as much as possible.Since the structure layer does not contain noise,this paper estimates the transmittance in the structure layer to avoid the influence of noise;the texture layer contains the texture information and noise of the image,therefore this paper denoises firstly on the texture layer.In order to preserve the better detail information,a binary mask and a weight matrix are introduced to fuse the texture layer and the structure layer in order to preserve the image details and suppress the noise.Experiments show that the proposed decomposition method can extract structural contour and background light information to low-frequency structure layer,and decompose noise and details into high-frequency texture layer.Experimental results show that the proposed enhancement method is superior to many existing underwater image enhancement schemes,and has achieved stable enhancement effect in underwater video enhancement.