In this paper, an automatic system to locate frontispieces in ancient scores is presented. The system is able to identify images containing two pages and separate them. Then, the algorithm performs a correction of the rotation. The next step is to delimit the area of interest which will be used along the next steps. After that, a mask containing the shape of staves is calculated to identify the areas with information different from the staff related musical information. With this, an in-determined amount of regions are extracted. Between these candidates, some restrictions are applied to obtain the final candidates to frontispieces. From the final candidates, the possible frontispieces are selected. The algorithm obtains good results and constitutes an important piece in a global OMR system.