Research On Underwater Image Enhancement And Restoration Algorithm

In recent years,with the decline of water,energy and mineral resources on land,human beings have gradually directed their exploration and development to the ocean.However,due to the special characteristics of marine environment,it is very difficult to observe and detect marine resources directly,and underwater images become a powerful tool to reflect the information of marine environment.Therefore,the analysis and processing of underwater images is an important research direction in the field of marine scientific research.In this context,the main work of this thesis has the following two points:To address the problems of color distortion and low contrast in underwater images,we propose an algorithm based on a LAB color space based color correction method and an adaptive look-up-table based contrast enhancement method.Firstly the algorithm performs adaptive color compensation on underwater images,determines the compensation parameters by calculating the degree of image color deviation,and compensates the attenuation of red channel with blue-green channel to prevent the image from red oversaturation after subsequent color correction operations.Then,the color-compensated images perform the color correction and contrast enhancement method proposed in this thesis,respectively.Finally,the color-corrected and contrast-enhanced images are fused using the image pyramid fusion method to achieve the purpose of complementary advantages.The experiments show that the proposed underwater image enhancement algorithm can effectively correct the color and improve the contrast.To recover the original underwater scene features,an end-to-end collaborative learning network is proposed for underwater image recovery based on an underwater image imaging model.The network uses a transmission map estimation network and a global ambient light estimation network to estimate the transmission map and global ambient light,respectively.Traditional image recovery algorithms usually estimate the transmission map and global ambient light separately,ignoring the interdependence between the transmission map and global ambient light.To settle this issue,a collaborative learning module is designed to enhance the information exchange between the transmisson map estimation network and the global ambient light estimation network.Experiments show that the underwater image recovery algorithm proposed in this thesis can effectively perform parameter estimation.