The first phase in optical music recognition (OMR) is staff-removal. The goal of this phase is to remove staffs and keep musical symbols with high precision, which is very important for next phases in OMR including segmentation and classification. In this paper a novel staff-removal method is proposed. In this method we used run-length coding to compress the input image and then the staff-lines are detected and removed using two common features, namely the thickness of lines and the space between them. Afterwards, The nonstaff objects such as long objects, dynamic wage and destroyed texts are removed using projection and morphological operations.