Klasifikasi Style Lukisan Menggunakan Self-Organizing Maps (SOM)

Andriyunanto, Reza (2017) Klasifikasi Style Lukisan Menggunakan Self-Organizing Maps (SOM). Undergraduate thesis, Institut Teknologi SepuluhNopember.

[img]
Preview
Text
5113100140-Undergraduate_Theses.pdf - Published Version

Download (5MB) | Preview

Abstract

Dalam pengenalan style citra lukisan, umumnya dilakukan secata subjective berdasarkan literasi yang ada. Style lukisan sendiri ada banyak, mulai dari abstrak sampai style modern. Untuk mengenali style lukisan seperti itu dibutuhkan fitur objective yang merepresentasikan setiap lukisan berdasarkan style. Tugas akhir ini mengusulkan sebuah metode klasifikasi menggunakan self-organizing maps untuk citra lukisan. Dilakukan ekstraksi fitur sebanyak 50 yang mewakili fitur lokal dan fitur global. Dimana fitur global mencakup fitur warna, sedangkan pada fitur lokal mencakup fitur warna dan fitur komposisi. Pada fitur lokal dibutuhkan proses untuk mendapatkan segmen pada citra, sehingga dilakukan segmentasi dengan menggunakan K-Means clustering lalu dilakukan filtering dengan median filter untuk memperbaiki hasil segmentasi. Hasil dari fitur tadi akan diolah dengan menggunakan self-organizing maps sehingga menghasilkan pengelompokan berdasarkan style lukisan. Ujicoba yang dilakukan pada tugas akhir ini menunjukan bahwa dengan learning rate 0.2 dan k pada K-Means ketika segmentasi 8 menghasilkan akurasi klasifikasi sebesar 53%. ================================================================= In the introduction of image painting style, generally done in a subjective based on existing literacy. Style painting itself there are many, ranging from abstract to modern style. To recognize the style of painting like that required feature objective that represents each painting based on style. This final project proposes a classification method using self-organizing maps for painting images. 50 feature extractions are performed that represent local features and global features. Where the global features include color features, while in local features include color features and composition features. On the local features required process to get the segment in the image, so it is done segmentation by using K-Means clustering then done filtering with median filter to improve the result of segmentation.The results of these features will be processed by using self-organizing maps to produce groupings based on painting style. The experiments performed on this final project showed that with learning rate of 0.2 and k on K-Means when segmentation 8 resulted in classification accuracy of 53% .

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Self-Organizing Map, painting style, K-Means clustering
Subjects: T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
Divisions: Faculty of Information Technology > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User: Reza Andriyunanto
Date Deposited: 28 Sep 2017 03:24
Last Modified: 06 Mar 2019 07:23
URI: http://repository.its.ac.id/id/eprint/44842

Actions (login required)

View Item View Item