การจำแนกฟันจากภาพเอกซเรย์โดยวิธีการเข้าคู่รูปแบบ

Abstract

This engineering project proposed a new method to segment a teeth from panoramic dental x-ray. Presently, dental information is widely used in many application because teeth features are distinguishable. So we make it useful by improving the program to segment a teeth from dental x-ray that will create more choices to get the primary screening before diagnosis by the expert. This research proposed a new method to segment a teeth from panoramic dental x-ray. It has 3 main sections including: Specify teeth area, Matching and Segmentation. Specify teeth area has 2 processes: 1. Converted to binary image using Otsu’s Threshold and 2. Define teeth area using Mahalanobis Distance. Matching step also has 2 processes: 1. Image Enhancement using Adaptive threshold and 2. Template Matching using Correlations. In the result shows performance of suitable templates using in Template Matching, get 4 sizes with mean accuracy is 60.75 % for single root and 60.83 % for double roots. The last section is Segmentation has 3 processes: 1. Plotted boxes merging, 2. K-mean Clustering and 3.Separating individual tooth. In the result shows performance of 4 boxes 3 boxes 2 boxes merging and 1 box from matching. An accuracy of this step has 48.91 per cent for double roots and 58.31% for single root so, we choose first 5detected teeth for single-rooted teeth and first 3detected teeth for double-rooted teeth and this step also uses detected area for declined template matching next step. Accuracy of declined template matching has 60.65% for single-rooted and 61.18% for double-rooted in approximately the results quite good for this project.