Seamless Image Fusion In Spatial Domain

Image fusion is a technology to synthesize the complementary information of multiple images acquired by two or more sensors for the same object or scene and fuses them into a composite image which has more accurate,stable and comprehensive information than a single source image.The goal of image fusion is to achieve the fine integration of complementary features through certain rules.It can be found that although many fusion algorithms can achieve good results in objective evaluation index,they are not satisfactory in subjective vision,which often makes the fusion results look particularly unnatural,and may even appear”block effect”.The main reason is that the gray value of the pixels changes sharply because of the poor convergence or connection at the edges of different features.Based on this,a seamless image fusion method in spatial domain is proposed in this paper,which aims to achieve seamless connection between different features and achieve a more visual fusion result.The method applies the weighting sum rules to complete image fusion in spatial domain.Therefore,how to obtain weights and how to optimize weights are directly related to the accuracy of seamless fusion,the core of this framework is to integrate Energy of Laplacian and Joint Bilateral Filter to achieve fusion.Firstly,a robust weight construction method is created.The specific process is as follows: Firstly,Structure Preserving Filter is used to preprocess the source images to eliminate the small-scale texture;Secondly,Improved Energy of Laplacian is used to detect the significant information and calculate the image gradient;Thirdly,Morphological Closed Operation is used to improve the gradient map;Fourthly,the initial weights are obtained by significant comparison and binarization operation,and it represents the most prominent feature pixels in the source image,but it cannot be directly applied to image fusion because of the edge sharpening.Secondly,Joint Bilateral Filter and an iteration process are used to optimize the initial weight map,a more natural final weight map with edges is obtained,it is necessary to ensure the accuracy of seamless connection between different features.Finally,the source image is processed directly by the weighting sum rules in spatial domain to get the final fusion result.In order to verify the superiority of the proposed method,three types of images and six typical image fusion methods are selected for comparative experiments,and quality and quantity analysis is given from subjective and objective evaluation.The fusion results show that the proposed method has good subjective visual effect.It can effectively solve the problem of the poor connection between different features and avoid the occurrence of pixel mutation,visual artifacts,halo phenomena,and also obtain higher objective evaluation index,which proves the superiority of the proposed method.