Study On The Restoration Of Underwater Image

Underwater images have been widely used in ocean energy exploration,marine environment monitoring and protection,as well as marine military and other fields.Due to the selective absorption of the water,traditional underwater image restoration algorithms which based on Jaffe-McGlamery simplified formation model have the disadvantages of high transmission and cannot estimate transmission of three channels respectively.In regard to these problems,this paper has proposed three different underwater image restoration algorithms.According to the influence of color distortion and color tone on dark channel prior in underwater image,this paper eliminates the backscattering after correcting the color distortion.Meanwhile,a fixed transmission for each channel is obtained by utilizing the ratio of the light scattering coefficient in water.Improving the estimation method for background light can also help to get a better quality of image restoration.The result of color constancy algorithms for the computation of the underwater image light color is constant in the global,which is contrary to attenuation state of light in water.To solve this problem,this paper proposes an underwater image restoration method based on local Shade of Gray algorithm.Assuming that the light is uniform and no attenuation in the local area,the color distribution pattern of light source reflecting light attenuation characteristics is obtained by calculating light color in each local region of image.Experimental results show that the proposed algorithm is simple,and can achieve a very good image restoration performance without refining transmission.Aiming at the shortcomings of image restoration method based on the Red Channel prior,this paper proposes an improved method.This method applies a new imaging model which is more suitable for the characteristics of light in water,and improves the accuracy of judging the existence of artificial light source.The attenuated ratio of the scene radiance and scattering ratio of background light are obtained respectively by using the Red Channel prior,image features and relationship between optical properties.Experimental results show that the improved algorithm has better visual effect than other state-of-the-art underwater image restoration methods.