Abstract: A method fused by Extended Fractal (EF) feature and local contract feature is proposed for target detection in SAR image. The paper mainly discussed the size sensitivity behavior for the EF feature and the second-order statistics for the EF feature in various target/clutter models, and concluded that the feature is also invariant to negative scalar multiplication of the image in the sense that a deep target-sized shadow can also be detected as well as bright target-sized objects. While the CFAR method only using the local contract information is not symmetric, it has a high false alarm in the strong clutter environment. Fusion of the two features provides an even lower false alarm rate when the targets can be detected. Experiments with real data show the effective of the fusion method.