Segmentasi Interaktif Berbasis Region Merging pada Citra Dental Cone Beam Computed Tomography

Firdaus, Fajar Maulana (2019) Segmentasi Interaktif Berbasis Region Merging pada Citra Dental Cone Beam Computed Tomography. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111540000171_Undergraduate_Theses.pdf Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Terdapat banyak sekali metode yang bisa digunakan dalam Image Segmentation. Secara umum, ada 3 kategori metode dalam image segmentation, yaitu manual segmentation, semi-automatic segmentation, dan automatic segmentation. Automatic Segmentation menggunakan tekstur, bentuk, ataupun warna yang ada dalam sebuah citra. Dalam beberapa kasus, seperti pada citra Cone-Beam Computed Tomography (CBCT), automatic image segmentation akan sulit untuk diimplementasikan, karena biasanya citra CBCT mengandung beberapa noise atau bintik-bintik kecil pada citra yang bisa menyebabkan kesalahan pada proses image segmentation menggunakan metode automatic segmentation. Selain itu, citra CBCT juga memiliki kontras yang rendah. Beberapa teknik semi-automatic segmentation dikembangkan. Metode ini menggabungkan manual segmentation dan juga automatic segmentation dengan menambahkan informasi dari pengguna pada proses segmentasi. Pengguna menambahkan informasi fitur pada citra untuk meningkatkan hasil dari image segmentation. Informasi yang diberikan kemudian digunakan pada proses segmentasi untuk memisahkan objek dengan background. Semi-automatic segmentation juga bisa disebut sebagai interactive image segmentation. Dalam tugas akhir ini, digunakan teknik semi-automatic segmentation dalam segmentasi citra CBCT. Pengguna menandai bagian dari objek dan juga background, sehingga nantinya akan terdapat 3 kelompok yang memisahkan citra, yaitu kelompok Object, Background, dan Non-Marked. Proses penggabungan region dilakukan dengan menghitung dan membandingkan jarak antara semua region yang tidak ditandai (Non-Marked) dengan Object dan Background. Hasil dari Tugas Akhir ini adalah nilai Misclassification Error (ME) sebesar 11,49%, Relative Foreground Area Error (RAE) sebesar 26,87%, Jaccard Coefficient sebesar 68,61%, dan Dice-Sorensen Coefficient sebesar 80,15% berdasarkan perbandingan hasil segmentasi dengan citra Ground Truth.
====================================================================================================================================
There are many methods that can be used in Image Segmentation. In general, there are 3 categories of methods in image segmentation, namely manual segmentation, semi-automatic segmentation, and automatic segmentation. Automatic Segmentation uses textures, shapes, or colors in an image. In some cases, such as in Cone-Beam Computed Tomography (CBCT) imagery, automatic image segmentation will be difficult to implement, because CBCT images usually contain some noise or small spots on the image that can cause errors in the image segmentation process using the automatic segmentation method. In addition, CBCT images also have low contrast. Several semi-automatic segmentation techniques were developed. This method combines manual segmentation and automatic segmentation by adding information from users to the segmentation process. The user adds feature information to the image to improve the results of image segmentation. The information provided is then used in the segmentation process to separate objects from the background. Semi-automatic segmentation can also be called interactive image segmentation. In this final project, a semi-automatic segmentation technique in segmenting CBCT images is used. The user marks the part of the object and also the background, so that later there will be 3 groups that separate the image, namely the Object, Background, and Non-Marked groups. The process of combining regions is done by counting and comparing the distance between all regions that are not marked (Non-Marked) with Object and Background. The results of this Final Project are the value of Misclassification Error (ME) are 11.49%, Relative Foreground Area Error (RAE) are 26.87%, Jaccard Coefficient are 68.61%, and Dice-Sorensen Coefficient are 80.15%, based on comparison of segmentation results and the Ground Truth image.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Segmentasi Citra Interaktif, Region Merging, Low Contrast, Dental Cone Beam Computed Tomography
Subjects:	R Medicine > RK Dentistry T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing
Divisions:	Faculty of Information and Communication Technology > Informatics > 55201-(S1) Undergraduate Thesis
Depositing User:	Fajar Maulana Firdaus
Date Deposited:	23 Jul 2026 08:40
Last Modified:	23 Jul 2026 08:40
URI:	http://repository.its.ac.id/id/eprint/66585

Actions (login required)

View Item