| **Input**: a blurred real-time image and a clear reference image |

| **Output**: the predicted matching position , the recovered image , and the estimated blur kernel |

| **Preparation**: construct the image dictionary from the reference image, the coordinate dictionary , the blur invariant feature , and dictionary ; |

| **Initialization**: initialize by solving sparse representation of w.r.t , and the restored image as ; |

| **for **** do** |

| Updating the blur kernel by solving equation (13); |

| Updating the recovered image by solving equation (15); |

| Updating the sparse coefficient by solving equation (21); |

| **end** |

| Predicting the matching position by equation (23); |