Sintesis Ekspresi Wajah Realistik Berbasis Feature-Point Cluster Menggunakan Radial Basis Function

Gunanto, Samuel Gandang (2018) Sintesis Ekspresi Wajah Realistik Berbasis Feature-Point Cluster Menggunakan Radial Basis Function. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

[img]
Preview
Text
07111160010007- Disertation.pdf - Accepted Version

Download (6MB) | Preview

Abstract

Meningkatnya permintaan produk animasi oleh rumah produksi dan stasiun televisi menuntut adanya perubahan yang signifikan di dalam proses produksi animasi. Penelitian animasi ekspresi pada wajah khususnya mengenai proses rigging dan pemindahan ekspresi semakin banyak. Pendekatan tradisional animasi ekspresi wajah sangat tergantung pada animator dalam pembuatan gerakan kunci dan rangkaian gerakan ekspresi wajah. Hal ini menyebabkan produksi animasi wajah untuk satu wajah tidak dapat digunakan ulang secara langsung untuk wajah lainnya karena kekhususannya tersebut. Oleh karena itu proses otomatisasi pembentukan area pembobotan pada model wajah 3D dengan pendekatan cluster berikut proses duplikasi gerak yang adaptif terhadap bentuk wajah untuk mempersingkat proses produksi animasi sangat penting. Prinsip animasi dipandang sebagai salah satu solusi dan panduan untuk pembuatan animasi gerak wajah yang ekspresif dan hidup. Sintesis ekspresi wajah realistik dapat dibuat dengan basis feature-point cluster menggunakan radial basis function. Otomatisasi pembentukan area gerak di wajah hasil proses clustering berdasarkan letak fitur titik dan proses retargeting menggunakan radial basis function untuk melakukan sintesis ekspresi wajah realistik merupakan kebaruan yang diangkat pada penelitian ini. Berdasarkan semua tahapan eksperimentasi yang dilakukan dapat disimpulkan bahwa sintesis ekspresi wajah realistik dengan basis feature-point cluster menggunakan radial basis function dapat diterapkan pada beragam model wajah 3D dan dapat secara adaptif peka terhadap bentuk wajah dari masing-masing model 3D yang memiliki jumlah fitur penanda yang sama. Hasil persepsi visual evaluasi penerapan sintesis ekspresi wajah realistik menunjukkan hasil ekspresi terkejut memiliki persentasi paling tinggi mudah dikenali, yaitu: 89,32%. Ekspresi senang: 84,63 %, ekspresi sedih: 77,32%, ekspresi marah: 76,64%, ekspresi jijik: 76,45%, serta ekspresi takut: 76,44%. Rerata persentase wajah mudah dikenali sebesar 80,13%. ================================================================================================================== The increasing demand of animated movies by production houses and television stations needs a significant change in the animation production process. Computer facial animation research on the process of rigging and expression transfer is growing. The traditional approach of facial animation is highly dependent on the animator in making the key and the sequence of facial expression movements. This causes the production of facial animation for one face can not be reused directly for the other face because of its uniqueness. Therefore, the process of automating the formation of weighted areas on 3D face model with cluster approach and adaptive motion transfer process to face shape is very important to shorten the production process of animation. The principle of animation is seen as one of the solutions and guidelines for the creation of animated facial expression expressively. The synthesis of realistic facial expression can be made on the basis of a feature-point cluster using a radial basis function. Automation process for formatting the motion area in the face by clustering process based on the location of the feature-point and retargeting process using radial basis function to perform synthesis of realistic facial expression is the novelty of this research. Based on all experimentation stages, it can be concluded that the synthesis of realistic facial expression based on a feature-point cluster using radial basis function can be applied to various 3D face models and can be adaptively sensitive to the facial shape of each 3D model which has the same number of marker features. The results of visual perception evaluation from the synthesis of realistic facial expression show that surprise expression has the highest percentage and easily recognizable, 89,32%. Happy expression: 84,63%, sad expression: 77,32%, angry expression: 76,64%, disgust expression: 76,45%, and a fear expression: 76,44%. The average percentage of faces is easily recognizable at 80,13%.

Item Type: Thesis (Doctoral)
Additional Information: RDE 621.367 Gun s
Uncontrolled Keywords: Sintesis ekspresi; pendekatan cluster; radial basis function; Expression Synthesis; cluster approach
Subjects: T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
T Technology > TR Photography > TR897.7 Computer animation
Divisions: Faculty of Electrical Technology > Electrical Engineering > 20001-(S3) PhD Thesis
Depositing User: Gunanto Samuel Gandang
Date Deposited: 03 May 2018 04:01
Last Modified: 24 Jun 2020 08:04
URI: http://repository.its.ac.id/id/eprint/51419

Actions (login required)

View Item View Item