The murals of the Tang tombs are important materials for studying the social life of the Tang Dynasty, which haveimportant conservation issues and historical research value. In order to preserve the tomb murals for as longer aspossible, it is necessary to restore the murals and accurately record the restoration location. However, the restoredmurals are difficult to discern directly with the naked eye. This paper proposes a method to reveal the restored areas, byextracting the main components of the spectral images of the mural with the Minimum Noise Fraction (MNF) Rotation.The location of the restored area is clearly observed from the main component. In this paper, we also proposed a newmethod to extract the sketches in Tang tomb murals. For the bands sensitive to the composition of the sketches, thesparsely constrained sparse non-negative matrix under-approximation method is used to decompose the optimalsketches composition, and then the sketches are automatically extracted based on the idea of layer superposition.Through the experiments on the mural paintings in the three tombs, the results demonstrated that the proposed methodcould effectively perceive the area of mural restoration and automatically extract the sketches accurately and clearly,while saving manpower.
展开▼