Underwater Image Enhancement Method Based On Gradient Domain

With the improvement of the imaging equipment's hardware and the reduction of the cost,the computer vision system has been more and more widely used in underwater scenes instead of artificial work.However,the prerequisite for underwater applications(such as marine exploration,environment monitoring and underwater objects recognition,etc.)to work well is that images captured by the equipment have high visibility and obvious features.Images captured in these underwater conditions are often severely degraded,which significantly reduces the performance of computer vision systems.According to the propagation characteristics of light,absorption and scattering are the mainly factors affecting the quality of underwater image.The absorption of light by water is wavelength dependent,which leads to color casts for images.Besides,the scattering caused by the underwater suspended particles leads to blur for images and low level of details,bringing great difficulties to subsequent applications based on visual perception.Therefore,it is particularly necessary to enhance the images visibility of complicated underwater conditions.Based on the comprehensive analysis of the latest underwater image enhancement algorithms and the development status at home and abroad,this dissertation makes a further study from the two aspects of complicated underwater conditions:color casts corrected and detail enhancement.Under these two conditions,a method for underwater image visibility enhancement based on gradient domain are proposed.The main contributions of this dissertation are introduced below.(1)According to underwater light attenuation law,the method of color correction based on white balance is applied for compensate the color casts and restore natural color of underwater images.(2)In order to improve the blur and detail loss of color corrected image,the detail enhancement based on the image gradient domain is applied to the image brightness channel.This method avoids the interference of image color information.Gradient fidelity term and two constraint terms are used to reconstruct brightness images according to energy optimization function.This step enhances the contrast of brightness images and obtains more detail information.(3)According to the degradation model,the transmission of image decreases with the increase of color attenuation degree.Utilized the attenuation amplitude from image,the transmission is used to compensate the color saturation to achieve more realistic color of underwater image.After the subjective evaluation and objective quantitative evaluation of large amount of recovered images,results produced by the proposed method are more satisfactory.When it comes to the extreme underwater condition,such as high turbidity images,advantage of our method is more obvious.