Abstract: In order to restore damaged radiography images, a new image inpainting algorithm for radiography has been proposed based on partial differential equations. This algorithm adopts p-Laplace operator of variational parameter into the curvature driven diffusions(CDD) model, and restores the damaged image using the improved model and the BSCB model by turns. By using the information along both tangent directions and gradient directions, the proposed algorithm can meet the connectivity principle of human vision perception and restore large damaged domain satisfactorily. It is proved that the new algorithm has better performance in convergence speed and inpainting quality compared with CDD model.