Research On Underwater Image Restoration Method Based On Imaging Mode

Due to the absorption of water body and the scattering effect of suspended particles,images captured underwater often suffer from quality degradation,which brings great trouble to the application of target segmentation,image recognition and scene reconstruction.Our proposed method belongs to the category of ‘underwater image restoration method’,which aims to recover the degraded image based on physical degradation model.The main work and contributions of this paper can be summarized as follows.To address the low contrast,color distortion,and poor visibility of underwater images,a novel underwater image restoration method is proposed based on scene depth estimation.First,the background light is estimated based on an improved quad-tree subdivision algorithm.Then,the scene depth is estimated based on the assumption that far scene regions are smoother and the intensity of the image gradient can be a rough estimation.Depending on the accuracy estimation of scene depth obtained by combing the oblique gradient operator(OGO)and underwater light attenuation prior(ULAP),the transmission maps can be further precisely acquired.After the background light and transmission maps are determined,the scene radiance can be finally restored based on the underwater image formation model.Experimental results demonstrate that the proposed method has a good performance on dehazing,color correction and contrast enhancement.To avoid the interference of foreground objects on background light estimation,we further propose a novel underwater image restoration method based on background segmentation and scene depth estimation.First,the degraded underwater image is divided into the foreground region and the background region according to the image gradient and color difference information.Accordingly,the background light is estimated in the background region.Then,the scene depth is estimated using multiple oblique gradient operator and attenuation difference among color channels.The transmission maps are obtained using the estimated scene depth map.After that,the scene radiance of foreground region is recovered based the underwater image formation model,and the background region is enhanced by performing histogram stretching in HSV color space.Finally,the foreground and background are fused with a weight map of the transition region to obtain the final restoration result.Experimental results show that the proposed method can estimate the background light and transmittance more accurately,and achieves a good performance in contrast enhancement,color correction and sharpness improvement.To solve the non-uniform illumination caused by artificial auxiliary light source,we extend underwater image formation model by introducing local background light,and present a novel underwater restoration method based on background extraction and underwater maximum reflection prior(UMRP).Considering that the UMRP is not suitable for underwater images with a large area of background,we first utilize three effective features including gradient,color difference,and area proportion to segment the background component and the foreground component.Then,we propose a simple but effective image prior,underwater maximum reflectance prior to estimate the varying underwater background light and transmission maps for foreground scene restoration.In addition,the background is enhanced by contrast enhancement.Experiment on real-world underwater images demonstrates that our method can not only increase contrast and remove haze effect,but also solve the non-uniform scattering effect caused by additional artificial illumination.