Remote sensing image matching via Harris detector and wavelet domain

Image matching is a hot issue in the analysis of the remote sensing data. The precision of the image matching have played an important pole on the application of the remote sensing image, such as three-dimensional modeling, data, pattern reorganization, panoramic reconstruction, superresolution, and phasic analysis. At present, the development of matching technology at home and abroad is mainly focus on the global variables. But from the geometric point of view, the process of shooting the image is a central projection, which forms the image to the different terrain from the different angle. Therefore, the image should have various deformation parameters in its local position. The key step within the process of image matching is the research on the estimation of the local feature parameters. In this paper, the authors introduce a matching algorithm that combine Harris detector with wavelet domain method. Our approach starts with the procedure of building a pyramid model on the remote sensing image. In the scale space, Harris detector was calculated from various scales under the constraint of wavelet coefficient (LH and HL) in the same level. Moreover, Taylor expansions algorithms are used to locate key point accurately in order to avoiding the phenomenon of location shifting in the process of extracting feature. At last, cross-correlation technology was used as a similarity to perform matching from the key point in the reference image to the sets of point in the querying image. Experimental results illustrate the good performance have been obtained in the precision and efficiency of the matching procedure.