Research Of Underwater Image Processing Based On Dark Channel Prior And Retinex

With the continuous development of underwater optical imaging technology,underwater images are widely used in underwater robot navigation,marine resource exploration,marine military and other engineering applications fields.However,due to the light absorption and scattering effect with the water medium,underwater images usually appear color distortion,low contrast,blurry detail,uneven illumination and so on.Based on thorough study of the physical properties of underwater optical transmission,this thesis presents two underwater image processing algorithms from two aspects: image enhancement and restoration.The specific work is as following.Aiming at the three color channels of R,G and B are independently processed by the traditional retinex enhancement algorithm,an improved multiscale retinex enhancement algorithm is proposed in this thesis.Firstly,only the image brightness component is processed by multi-scale retinex enhancement algorithm,and reflected image is decomposed from it.Secondly,the gray value of each channel is readjusted based on the gray value ratio between color channel of the original image and the brightness component.Finally,the adjusted grayscale values of each channel are linearly stretched based on the normal distribution.Experimental results show that the improved multi-scale retinex enhancement algorithm can effectively remove the blue-green hues of the underwater images and improve the image contrast.The subjective visual effects of the enhanced underwater images are more suitable with human visual system.Based on the Jaffe-McGlamery underwater optical imaging model,a simplified underwater imaging model is proposed.Firstly,the R-channel total attenuation coefficient transmittance is estimated according to the red channel prior.Considering mathematical relationship between background light and total attenuation coefficient and scattering coefficient,the blue and green channel total attenuation coefficient transmittance are calculated from the red channel,and the target scene reflection that removes the background light is restored.Secondly,according to the mathematical relationship between optical transfer function and transmittance,inverse filtering is applied for removing the forward scatter component so as to improve image detail clarity.Finally,according to the variational retinex,the reflection and the illumination images are decomposed from the luminance component in the Lab color space.In order to adjust the uneven lighting,the illumination image is gamma corrected.The proposed and four state-of-the-art underwater image processing algorithms are used to process several images captured in various underwater environments.From the two aspects of subjective visual evaluation and objective evaluation index,the results of five algorithms are analyzed.The experimental results prove that the proposed algorithm can effectively remove the effects of light absorption and scattering attenuation,which increases the clarity of the target object.And the proposed algorithm has superiority for balancing the color,contrast,and saturation of the images,the visual effects are more similar to images captured in natural scene.