Automatic registration for ASAR and TM images based on region features

摘要

Aimed at lower-precision and time-consuming shortcomings in registration for SAR and optical remote sensing images, a region feature based registration method was proposed. A modified Lee filter was employed to suppress speckles in SAR image firstly. Than, image segmentation based on region growing restricted by the edge was used to extract closed regions in SAR and optical images as invariable features. After that precise feature matching was guided by a cost function, which considering region area, perimeter, coordinate of center point and so on. After cross correlation matching, the center points of the matched region were taken as registration control points (RCPs). Experimentation was performed by registering Envisat ASAR and Landsat TM images. The results show that the whole workflow could be accomplished automatically and the root mean square error (RMSE) of the RCPs got by this method is lower than 1 pixel.