Remote Sensing Image Fusion Algorithm Based On Joint Filtering Of Mutual-structure And Saliency Detection

With the wide application of computer technology,image technology has made great progress,such as in national defense,environment,agriculture and forestry,water conservancy and other important areas.The traditional single type of remote sensing data can not meet the growing research needs of people.We need to simultaneously and synchronously use hyperspectral and high spatial resolution information.Therefore,the multi-source remote sensing image method has been paid more and more attention by scholars.In practical application,hyperspectral and low spatial resolution multispectral images(MS)are mainly used to accurately identify land cover types,marine spectral characteristics analysis,geological exploration,surface vegetation research,military target detection and recognition and other accurate classification directions.Pan with high spatial and low spectral resolution is mainly used to describe the texture and shape of features,such as environmental monitoring,land use,urban planning,resource management and so on.In both cases,the single image can not provide enough information.Therefore,remote sensing image fusion is essential.Multi-source image can enrich visual information and reduce uncertainty.Remote sensing image fusion makes full use of the difference and complementarity of multi-source remote sensing image.The ultimate goal is to fuse MS image and PAN image to generate fusion image with high spatial resolution and hyperspectral resolution.Compared with the information contained in any source image,the obtained fusion image is more comprehensive,which improves the reliability of data,improves the possibility of efficient interpretation of remote sensing image,deepens the application potential and value of this field,and is of great significance to the development of national defense construction,economic development and national life in China.First of all,the current research status of remote sensing image fusion,hierarchical division and algorithm classification are briefly described,and the relevant concepts and knowledge of the algorithm are introduced.In addition,the evaluation methods and indexes of fusion image are also introduced.In this paper,we present a remote sensing image fusion method based on the mutual structure of joint filtering(MSJF)and significance detection.Our method uses joint filtering to extract high frequency and low frequency from source image.The most outstanding advantage of the joint filtering method is that it can effectively reduce the artifacts caused by the small differences between the source images.In the early process of filtering and decomposing,this method associates two images instead of filtering them separately,which makes the decomposing effect better.In addition,in the process of calculating the low-frequency fusion weights,we use the significance detection method based on human eyes to judge the image quality,so that the visual effect of the fusion results is better;for high-frequency fusion,the extended Laplace correction and(esml)method makes the fusion more detailed,because compared with the traditional SML,the calculation of diagonal direction is increased.Our method overcomes the shortcomings of some traditional methods,such as MRA based method has been flawed in spatial contrast details;SML may produce artifacts.In this paper,different satellite data(Landsat 8,IKONOS,QuickBird)are used to verify and compare the proposed method and five comparison algorithms,and the fusion results and data comparison results are given.In this paper,six evaluation indexes are selected,and the single band and combined band are tested respectively.Through the visual chart comparison,it shows that the algorithm effectively retains the structure information and texture information of the image,improves the clarity of the image and has many advantages in the subjective and objective evaluation.