Generating automatic marker based on combined directional images from frequency domain for Dental Panoramic Radiograph Segmentation

• Published in: 2017 Second International Conference on Informatics and Computing (ICIC) pp. 1 - 6
• Main Authors: Rahma, Amelia Sahira, Bisono, Eva Firdayanti, Arifin, Agus Zainal, Navastara, Dini Adni, Indraswari, Rarasmaya
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2017
• Subjects: automatic marker; dental panoramic radiograph; Dentistry; directional image; frequency domain; Frequency-domain analysis; Gabor filters; Histograms; Image segmentation; Radiography
• DOI: 10.1109/IAC.2017.8280627

Dental Panoramic Radiograph is often used for age identification and forensic examination in medical evaluations. Like most medical images, it has low contrast and high noise. Hence for facilitating medical workers in analyze the image, it requires segmentation process to separate foreground and background areas. Conventional image segmentation technique used markers to determine the object areas manually. Several researches that develop automatic marker for image segmentation based on spatial domain have been conducted. However, they are unable to provide satisfactory segmentation result, because it relies on pixel processing and not effective in images with high spatial size. Therefore, new automatic marker determination method is needed to overcome this problem. This paper proposed a new method for generating automatic marker based on combined directional images from frequency domain for Dental Panoramic Radiograph segmentation. This method creates directional images that separate the dental image into vertical and horizontal direction. Directional images are combined and used as inputs in automatic marker generation. Then the automatic marker can be applied on image segmentation technique, such as Region Growing algorithm. Experimental results on 40 dental images showed that the use of proposed automatic marker yields better results than the marker using vertical or horizontal directional image separately. With the best value of Misclassification Error (ME), Relative foreground Area Error (RAE) and Accuracy are 7.8%, 0.2% and 92.2% respectively. Automatic marker using combined directional images has stability for handling the segmentation of dental images that tend to have vertical and horizontal direction.