Research On SAR And Multispectral Image Fusion Algorithm For Feature Enhancement

Multi-source remote sensing image fusion refers to the process of integrating multiple remote sensing images of different sensors in the same area and synthesizing a new image by a specific technology.Because of the different imaging mechanism,synthetic aperture radar and multispectral images have great differences.SAR images have strong penetrating power and are not affected by bad weather.SAR images can be captured all day and represent rich surface texture and structure features.Multispectral images contain abundant spectral information,which can effectively identify different types of ground information and have superior visual effect.Therefore,the fusion of SAR and multispectral images can realize the complementary advantages of multi-source images and expand the application range of images.Based on the research background and significance of SAR and multispectral image fusion,this thesis summarizes the current research status and shortcomings of fusion algorithms at home and abroad.And the SAR and multispectral image fusion algorithm has been deeply studied.The main work includes the following two aspects:（1）A fusion algorithm for SAR and multispectral images based on phase congruence and pulse coupled neural network was proposed to solve the problems of spatial detail loss in fused images.In this thesis,intensity-hue-saturation and non-subsampled shearlet transform are used to decompose and reconstruct the image.In low frequency,the fusion rule based on phase consistency is adopted,which not only extracts the feature information of low-frequency image,but also avoids the loss of spectral information.PCNN fusion rules are used in high frequency to extract texture,edge and other details of the image effectively.（2）In order to improve the ability of local detail information extraction and image contrast,SAR and multispectral image fusion algorithm based on dual-tree complex wavelet transform and guided filtering is proposed.Based on IHS transform and NSST transform,this algorithm uses double tree complex wavelet transform to decompose the low frequency coefficient of NSST twice and fully extracts the details of low frequency.In the low frequency of DTCWT,the improved fusion rule of weighted average and guided filter is used to enhance the image contrast effectively.PCNN fusion rules are used in both DTCWT high frequency and NSST high frequency to fully retain the details of the image.In this thesis,three groups of Sentinel-1 SAR images and Landsat-8 multispectral images are used to conduct experiments.The results show that the proposed algorithm is significantly improved in both subjective and objective evaluation,and the fused image not only has high fidelity in spectral and spatial information,but also has enhanced contrast and clarity of the image.