Gunawan, Wawan (2017) Strategi Region Merging Berdasarkan Pengukuran Fuzzy Similarity pada Segmentasi Citra. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
5115201001-Master-Theses.pdf - Published Version Download (5MB) | Preview |
Abstract
Metode segmentasi citra semi otomatis dilakukan dengan cara membagi
citra menjadi beberapa region berdasarkan nilai kemiripan antar fitur-fiturnya.
Kemudian pengguna memberikan tanda pada beberapa region sebagai sample dari
region objek dan background. Selanjutnya sample region tersebut digunakan pada
proses region merging terhadap region yang belum ditandai berdasarkan nilai
kemiripannya. Beberapa region pada citra memiliki nilai informasi yang tidak
merata, seperti blurred contours, soft color shades, dan brightness. Region tersebut
pada penelitian ini kita sebut sebagai ambiguous region. Ambiguous region
menimbulkan permasalahan pada proses region merging dikarenakan region
tersebut memiliki dua nilai informasi yaitu sebagai objek dan background. Hal
tersebut dapat menimbulkan kesalahan dalam proses segmentasi.
Pada penelitian ini diusulkan strategi region merging baru berdasarkan
pengukuran fuzzy similarity pada segmentasi citra. Metode yang diusulkan
memiliki empat tahapan, tahap pertama adalah region splitting yang digunakan
untuk mendapatkan intial segmentasi. Tahap kedua adalah penandaan manual yang
dilakukan oleh pengguna untuk menandai sample dari region objek dan background
(user marking). Tahap ketiga adalah initial fuzzy region untuk mendapat inisial seed
background dan objek. Tahap terakhir adalah proses region merging menggunakan
pengukuran fuzzy similarity dengan memperhitungkan intensitas gray level dan
fungsi keaangotaan. Berdasarkan hasil uji coba metode yang diusulkan berhasil
melakukan segmentasi pada citra natural dan citra gigi dengan rata-rata nilai
misclassification error 1.96% untuk citra natural dan 5.47 % untuk citra gigi. Selain
itu metode yang diusulkan dapat menghasilkan segmentasi yang lebih akurat
dibandingkan dengan metode MSRM, Global FSM, dan Semi FSM.
=======================================================================================
Semi-automatic method of image segmentation can be done by dividing
the image into multiple regions based on the similarity between its features. Then
the user gives marks on several regions as a sample of the object region and
background region. Furthermore, the sample used in the process of region merging
between non-marker regions based on their similarity. Some regions of the image
have an unbalance information, such as blurred contours, soft color shades, and
brightness. We call those regions as ambiguous region. Ambiguous region cause
problems during the process of merging because that region has double information
as object and background. This can lead to segmentation error.
Therefore, we proposed new region merging strategy based on fuzzy
similarity measurement on image segmentation. The proposed method has four
stages; the first stage is region splitting used to get the initial segmentation. The
second stage is manual marking by the user to get a sample of the object region and
background. The third stage is determining the initial fuzzy region to receive initial
seed background and object. The last stage is the process of merging region against
non-marker region by determining the optimal threshold to the cluster background
region and object region using fuzzy similarity measurement taking into account
the gray level intensity and membership function. The proposed method is expected
to optimize image segmentation result than other region merging methods.
Experimental results demonstrated that the proposed method can be done
segmentation for natural and teeth image with the average value of misclassification
error (ME) 1.96% and 5.47% respectively. The proposed method can give accurate
segmentation result compared with MSRM, Global FSM, and Semi FSM.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | ambiguous region; pengukuran fuzzy similarity; segmentasi citra; strategi region merging; ambiguous region; fuzzy similarity measurement; image segmentation; region merging strategy. |
| Subjects: | Q Science > QA Mathematics > QA248_Fuzzy Sets |
| Divisions: | Faculty of Information Technology > Informatics Engineering > 55101-(S2) Master Thesis |
| Depositing User: | Wawan Gunawan |
| Date Deposited: | 20 Mar 2017 03:43 |
| Last Modified: | 05 Mar 2019 06:09 |
| URI: | http://repository.its.ac.id/id/eprint/1990 |
Actions (login required)
 |
View Item |