Segmentation is challenging process in medical images especially on dental x-ray images. Level set methods have best result on medical and dental image segmentation. Initial Contour (IC) is the essential step in level set methods to initialize the efficient process. However, the main issue with IC is how to generate the automatic technique in order to reduce the human interaction and produce accurate result. In this paper a new region-based technique for IC generation, is proposed to generate the most suitable IC. We have utilized the statistical and morphological information inside and outside the contour to establish a region-based map function. This function is able to find the suitable IC on images to perform by level set methods. Experiments on dental x-ray images demonstrate the robustness of segmentation process using proposed method even on noisy images and with weak boundary. Furthermore, computational cost of segmentation process is reduced.