Abstract:
This research proposes a method to recognize a tooth in dental X-ray images using tooth Hus moment invariants feature. The proposed tooth recognition method consists of three steps that are image enhancement, image segmentation and feature matching. Histogram equalization and noise reduction with median filter are used to enhance the input image. The teeth are automatic segmented by with Otsus thresholding method. The last step is to calculate Hus moment invariants that are used as the tooths feature and match with those of the reference teeth using Euclidian distance. The least distance implies the best matched result. Experiments are done with molars and premolars in the panoramic films of orthodontic patients taken at different time. From the experiments, the efficiency of the proposed method for recognizing the correct patient and the correct tooth was 26.67 % and recognizing the correct tooth only was 53.33%. For the manual segment image set, the results were 55.55% and 91.11% respectively. Hence, it can be seen that the accuracy of segmentation step has high impact to the recognition rates and that Hus moment invariants and Euclidean distance cannot give good recognition rate for recognizing the correct patient and the correct tooth.