Feature-based automatic registration for SAR and optical remote sensing images

In the registration for SAR and optical images, traditional methods are limited by SAR speckle noise and the accuracies of selected registration control points (RCP). A new registration method based on region features is presented. The speckle noises in SAR image are suppressed first and then image segmentation is used to extract close regions as invariable features. Meanwhile, the close contours in optical image are extracted by edge extraction. Next, the precise feature matching is made by chain-code correlation and registration accuracy of sub-pixel is obtained. Experimental results show that the method can extract RCPs with high accuracy and accomplish the automatic registration for multi-sensor images.