It is very difficult to extract the shoreline from SAR images by using a simple threshold technique for the presence of speckles. The applications of many coastline detection algorithms based on complex mathematic models are limited for their huge time consuming. A fast method is presented in this paper based on the idea of seeds growing. First, it locates the initial seeds region automatically based on statistic information of pixel values. Then, the initial mean M and threshold T are calculated. Next, seeds-growing is performed based on the M and T which are updated continuously. After the seeds-growing is ended, the location of the shoreline is extracted from the connected sea region by a traditional contour tracing algorithm. Experimental results using real SAR images indicate that it is an effective and real-time algorithm.