Thesis/Dissertation: Bc. Andrej Krejčír: GPU Implementation of Generalized Chan-Vese Model Minimization Using Graph Cuts
Master's thesis
GPU Implementation of Generalized Chan-Vese Model Minimization Using Graph Cuts
GPU implementace zobecněného Chan-Vese modelu minimalizovaného pomocí grafových řezů
Abstract
Cieľom tejto práce je implementovať a vyhodnotiť zovšeobecnený Chan-Vese model pre segmentáciu obrazu, minimalizovaný pomocou grafových rezov na GPU. Zovšeobecnenie je založené ma použití polynomiálneho modelu regiónov namiesto konštantných intenzít. Práca vysvetľuje detaily modelu a jeho implementácie. Presnosť výsledkov je porovnaná s pôvodným Chan-Vese modelom. Implementovaná GPU metóda pre nájdenie …more
Abstract
The aim of this thesis is to implement and evaluate a generalized Chan-Vese model for image segmentation, which is minimized using graph-cuts on the GPU. It is generalized by using polynomial region models instead of a constant function. The details of the model and the implementation are explained. The accuracy of results is compared to the original Chan-Vese model, and the performance and memory usage of the GPU graph-cut method are compared to existing CPU and GPU implementations.
Thesis description
Literatura:
- [1] S. Mukherjee, S. T. Acton. Region based segmentation in presence of intensity inhomogeneity using Legendre polynomials. IEEE Signal Processing Letters, 22(3):298-302, 2015.
4/1/2017 16:20, doc. RNDr. Martin Maška, Ph.D., UČO 60734
ČSN ISO 690-compliant citation record
Theses on a related topic
List of theses with an identical keyword.
-
Fast Level Set-Like Algorithms in Fluorescence Microscopy
doc. RNDr. Martin Maška, Ph.D., UČO 60734 -
Tool for data pre-processing and iterative learning of neural networks
Ing. Kristián Malák -
GPU Implementation of Piecewise Affine Mumford-Shah Model
Mgr. Bruno Čičmanec -
Terrain Rendering Using GPU
Mgr. Marek Mauder -
Design and implementation of arithmetic coder for CUDA platform
RNDr. Vít Rusňák, Ph.D., UČO 172757 -
Understanding the Sequence Preference of Recurrent RNA Building Blocks using Quantum Chemistry: The Itrastrand RNA Dinucleotide Platform
RNDr. Mgr. et Mgr. Arnošt Mládek, Ph.D. -
Methods for accelerating parallel algorithms using GPUs
Mgr. Jakub Kremláček, UČO 410131 -
Parallelization of brute force attack on MD5 hash algorithm in FPGA
Mgr. Maruthi Gillela




