A model of parallel pair is proposed to overcome the limitation of the traditional one and it is the kernel of the designed algorithm for parallel feature extraction in Synthetic Aperture Radar (SAR) images. After the original SAR image is filtered to suppress the speckle noise, Ratio of Exponentially Weighted Averages (ROEWA) operator which has constant false alarm rate is used to detect edge. Then the approach of primitive parallel line segment extraction is carried out. Finally, the breakpoints are connected by means of the heuristic link idea. The experiment results show that the proposed algorithm can extract the parallel characteristic of SAR images efficiently, and it can be used for auto target recognition, such as roads, airports and rivers.