Research On Underwater Optical Image Quality Perception And Restoration Algorithm

Underwater optical imaging technology is an important means to obtain marine information and is widely used in many important areas,including marine resource development and exploration,environmental monitoring and protection,topographical survey,cruise guidance,marine military as well as national defense construction.However,under the complex marine environment,because light has serious attenuation and harsh scattering problems,underwater images often present problems such as low contrast,blurred details,color attenuation,and low illumination,which cause great difficulties for further analysis and processing through computer vision.Therefore,it is extremely important to use underwater image processing technology to meet subsequent image analysis tasks.We mainly focus on two aspects of underwater image quality perception and processing,combining traditional algorithms with deep learning algorithms to process underwater images for different degradation features in this paper.We can obtain higher quality and more effective underwater visual information.For the purpose of collecting underwater images with better visual quality at the front end,firstly,based on visual saliency,an underwater image composition quality perception algorithm is proposed to guide the underwater vehicles to obtain better composition scene.Then we propose an assessment algorithm based on visual saliency for general quality of underwater image(VS-UIQA)to measure the overall quality,which contains underwater image attribute metrics:brightness contrast index based on exposure type(ET-UIConA),color difference index based on visual saliency(VS-UICA),and sharpness index based on visual saliency(VS-UISA).Each indicator can individually assess the degradation degree of the corresponding problem of underwater image.Finally,the statistical characteristics of diverse underwater images which have different degeneration features are discussed,which provide a theoretical basis for subsequent guidance of underwater image processing.The proposed algorithm has better performance compared with other several algorithms.Aiming at the color attenuation problem for underwater image,based on the color statistical results of different degraded underwater images,corresponding color compensation method is used,and white balance processing based on the color constancy theory is performed to realize color correction of underwater optical images.The algorithm in this paper can deal with various underwater degradation problems such as turbidity,blue-green color cast and low illumination.Aiming at the scattering problem in underwater image,we propose a processing algorithm based on generation adversarial network driven by dual image wavelet fusion algorithm(DIWF-GAN).The proposed underwater image quality perception indicators are used to guide the improving of image visibility quality.Finally,we perform the simulation data sets,color card data sets as well as real underwater image data sets to test the algorithm performance.Experiment results prove that the proposed algorithm can realize functions including low-light enhancement,color correction,and sharpening for underwater images with different degradation features.