Research On Remote Sensing Image Mosaic Algorithm

In the application of remote sensing images, in order to obtain a greater vision, betterhandle, analysis, research and interpret the information of remote sensing image, two ormore remote sensing images which have overlapping area need to be spliced one image.There is a very broad application field of the remote sensing image stitching technology,like vegetation resources investigation, land water resources investigation, environmentalmonitoring, marine resources investigation, land resources investigation and so on.Therefore, the remote sensing image stitching technology has become one of the basictechniques of macroscopic rule research and regional monitoring using remote sensing ingeography science.This paper first introduce the applications and research statues of the Image Stitchingtechnology at home and abroad. Then introduce the key technology like image registrationand image fusion detailly, analyzes and summarizes their main methods, properties andshortcomings. The image registration algorithm based on feature points and the imagefusion technology based on pixels are presented importantly.Firstly, SURF is used to achieve the extraction of feature points. Secondly, we use theratio of the the nearest neighbor and the second nearest neighbor to match the interestpoints, and we remove the mismatching points using RANSAC. Lastly, the fusionalgorithm based on Shearlet is used to achieve the fusion image, as the fusion rules is thecure of the image fusion, and it will directly affect the quality of image fusion, we outlinedthe rules of the low and high frequency emphatically, the weighted average algorithm isused in low frequency coefficients to achieve the smooth stitching results, and the regionalenergy is used in high frequency coefficients. We use the subjective and objectiveevaluation criteria to evaluate quality of the mosaic image. Experiments show that themosaic image using the algorithm this paper proposed get better result.