Study of the Automatic Matching Method for Optical and SAR Image

For the problem that existing methods produced in optical and SAR image matching for example false match rate arising and low efficiency due to the imaging mechanism, this paper presents a method that using the centroid of two matched surface feature as the homologous point. First, lee-sigma method was used for SAR image speckle suppression. Second, extract the edge of two images by canny operator, then the dilation and closed operations of morphology to be closed the edge map, and the surface feature were extracted using the eight neighborhood growth algorithm. Final, cost function was used for relevant feature matching in two images, at the same time the centroid was extracted as homologous point. Paper with Landsat-7 and Envisat-1 image for experiment, results show that this method could reduce the mismatch rate, improve efficiency, and contributed to achieve automatic matching, and the precision reached sub-pixel level.