We present a method of extending the dynamic range of an picture with two different exposure images. Since pictures under different exposure times show different scene dynamic ranges, if we make use of the visible information of each different exposure image, we can recover a high dynamic range image which contains what cannot be visualized by an auto-exposure time.In this method, we did not follow the ordinary fusion methods which were based on multi-layered or radiance map based approach of multiple images. Instead, we used the filling method of undesirable regions of an auto-exposure picture, preserving its overall image quality.Our filling method uses the gradient information of the visible regions and recovers the saturated regions by the energy minimization approach. The results show that our fusion method is simple and practical to make a HDR image, if we have two proper exposure images.