During lifetime, teeth are exposed to many effects like abrasion, loss and dental treatments. These effects along with natural shapes of teeth form a unique dental frame which contains useful attributes to be used for human identification. Today, there exist automated dental identification systems which are used by forensics of law departments. These systems need to extract dental structures like teeth or roots prior to further analysis. So far, in several studies, much effort has been paid for this task. However, there still exist core problems like automated detection of region of interest (ROI) and segmentation in panoramic dental radiographs with missing teeth. This study aims to present a tool that can be employed to overcome these issues. Unlike previous works, the proposed methodology takes advantage of discrete wavelet transform for more accurate localization of ROI and polynomial regression to form a smooth border, separating upper and lower jaws even in case of absent teeth. Results indicate that the proposed approach can be effectively used for teeth segmentation and root apex detection.