KOMPUTASI PARALEL PADA METODE DISTANCE REGULARIZED LEVEL SET EVOLUTION (DRLSE) UNTUK SEGMENTASI CITRA MEDIS

RIANTO, INDRA (2014) KOMPUTASI PARALEL PADA METODE DISTANCE REGULARIZED LEVEL SET EVOLUTION (DRLSE) UNTUK SEGMENTASI CITRA MEDIS. S2 thesis, UAJY.

[img]
Preview
Text (Halaman Judul)
0MIF01902.pdf

Download (959kB) | Preview
[img]
Preview
Text (Bab I)
1MIF01902.pdf

Download (109kB) | Preview
[img]
Preview
Text (Bab II)
2MIF01902.pdf

Download (163kB) | Preview
[img] Text (Bab III)
3MIF01902.pdf
Restricted to Registered users only

Download (330kB)
[img] Text (Bab IV)
4MIF01902.pdf
Restricted to Registered users only

Download (659kB)
[img]
Preview
Text (Bab V)
5MIF01902.pdf

Download (255kB) | Preview

Abstract

Medical image are considered important in medical world. Medical image can be used for surgery preparation and simulation, radiotherapy planning, and for monitoring wound development. Analising medical image is not an easy task, expertise and experience are needed to do this task. Even with expertise doing this task can take an amount of time. In this case computer is used to analise medical image, by using image processing technique, image segmentation. Medical image segmentation can be done using Distance Regularization Level Set Evolution (DRLSE) method. This method were developed from the level set method, which is used to eliminate the irregularities from level set method. By using this method, medical image segmentation can be done, but long processing time occur when using large files. To accelerate segmentation time, computation that were done sequentially are replaced with parallel computation by using Compute Unified Device Architecture (CUDA), which is developed by NVIDIA enabling access to Graphic Processing Unit (GPU). Using CUDA, the process of medical image segmentation using DRLSE can be accelerated. Images used are in 64x64, 128x128, 256x256, 512x512, and 1024x1024 sizes. Segementation time on 64x64 image shows no significant time difference between CPU and GPU. On larger files, computation time on each processing unit show different result. Segmentation time on 1024x1024 image, GPU can process image seven times faster than CPU.

Item Type: Thesis (S2)
Uncontrolled Keywords: NVIDIA CUDA, Parallel Computation, DRLSE
Subjects: Magister Teknik Informatika > Soft Computing
Divisions: Pasca Sarjana > Magister Teknik Informatika
Depositing User: Editor UAJY
Date Deposited: 16 Jun 2014 12:17
Last Modified: 16 Jun 2014 12:17
URI: http://e-journal.uajy.ac.id/id/eprint/5269

Actions (login required)

View Item View Item