Research On Underwater Image Enhancement Algorithm Of Multi-feature Fusion

The ocean is rich in resources and is an important strategic foundation for China to build a strong ocean state.As the main carrier of marine information,underwater images are widely used in areas such as resource development and environmental protection,and are of great importance to the national economy and national defense security.However,underwater images suffer from many problems such as color distortion and haze,making the image quality degraded.Therefore,it is crucial to propose effective methods for underwater image enhancement.Starting from the research hotspot of underwater image fusion strategy,this paper improves the visual quality of underwater images and makes the processed images more in line with human visual characteristics and the needs of marine engineering.The main research contents are as follows:(1)Aiming at the comprehensive effects of color distortion,non-uniform illumination,and low contrast of underwater images,an underwater image enhancement based on dominant feature image fusion is proposed.First,the dark channel prior is improved to correct color casts by proposing a strongly robust background light calculation method.Second,the CLAHE-HF and AGCWD algorithms are used for the color-corrected images,respectively,and the associated weight maps are defined to retain the dominant features of each image.Finally,the enhanced images are obtained by multi-scale fusion.The experimental results show that the fused images retain the characteristics of the improved dark channel prior algorithm that can remove uneven turbidity and achieve color balance,and also retain the advantages of the CLAHE-HF and AGCWD algorithms in terms of brightness adjustment and contrast enhancement.The fused image has high color fidelity,rich image details,and good dehazing effect.(2)Aiming at the problem of underwater haze and color distortion,an underwater image enhancement based on fusion of restored structure and enhanced texture is proposed.First,the image is decomposed into structure and texture layers by a relative total variance model.Second,the red-blue dark channel prior is proposed to be applied to the structure layer to achieve color equalization while dehazing.Then,the gradient smoothing method is proposed for enhancing the effective details in the texture layer.Finally,the restored structure layer is fused with the enhanced texture layer to obtain a clear underwater image.The method combines the advantages of restoration and enhancement techniques and is no longer limited to solving a single problem such as haze phenomena,achieving good results in complex underwater environments.The experimental results show that the method outperforms the comparison algorithm in all evaluation indexes,so it can be effectively used in marine engineering.(3)The proposed algorithms are compared with 12 novel or classical underwater image enhancement algorithms in recent years,and the effectiveness of both algorithms in underwater image enhancement is fully verified.Several representative datasets are used for comparison,and a comprehensive and objective experimental analysis is performed using seven evaluation metrics.The experimental results show that the proposed algorithms have excellent clarity performance compared with each comparative algorithm,in which the UCIQE measurement values of the UIEB dataset are improved by 20.12% and 13.86%,respectively.In addition,the proposed algorithm has been verified by extended applications to achieve stable enhancement results in several underwater applications as well.There are 50 figures,26 tables and 74 references in this thesis.