Research On Color Restoration And Enhancement Algorithm Of Underwater Image

As one of the necessary means of human underwater detection,underwater imaging technology is widely used.However,due to the influence of light absorption and scattering by water body and suspended objects in water,underwater image will produce color distortion,blurred image,low contrast,uneven illumination and other problems,which further hamper the observation of underwater environment and the extraction of underwater object features.Based on the research of underwater image imaging characteristics and imaging models,this paper proposes three underwater image enhancement and restoration algorithms,which can enhance image contrast and brightness,protect and enhance important information features of image,remove the influence of underwater light scattering,correct underwater color offset and improve image quality.The attenuation coefficients of different wavelengths of light in the underwater environment are calculated according to the scattering coefficients of light.Inspired by dynamic threshold algorithm,the image from RGB color space to YCb Cr color space,the attenuation coefficient is calculated by using three-channel attenuation after the offset parameter matrix and correction should be mapped value of Y,Cb,Cr,find images "white dot",so as to calculate each color channel gain,in addition to the underwater color offset,the underwater image blue-green tone conversion to normal white color tone of the atmosphere.Based on color constancy,Retinex algorithm can restore the image color and enhance the object features in the image.However,Retinex algorithm based on strong coupling RGB space will produce color distortion when processing underwater images.In view of the above situation,the brightness information in YCb Cr with low coupling is used to judge the ambient light,and then the bilateral filter is used to replace the Gaussian filter to protect the high frequency edge.Finally,the ambient light of each channel is solved respectively based on the underwater attenuation ability of each channel.The contrast of underwater image is improved,the object features and the information in the image are enhanced,and the color bias problem is corrected.The dark channel prior based on physical imaging model has always had a good restoration effect on "fog" images,and can also be applied to underwater image restoration.However,in the underwater image restoration,the algorithm based on the dark channel prior is insufficient in the calculation efficiency and the estimation of the ambient background light.An improved algorithm is proposed,which corrects the color of the image first,then uses interpolation between three channels and gradient in the pixel domain to estimate the accurate global background light,and uses down-sampling and fast nuclear expansion etching algorithm to calculate the dark channel and its transmittance map.On the basis of accurately restoring the underwater image and improving the image quality,the computing efficiency is greatly improved and the consumption of computing resources is reduced.