GPGPU Acceleration of Active Contour Segmentation

OData support
Supervisor:
Lois László Dr.
Department of Networked Systems and Services

Nowadays graphic cards have big potential in arithmetic operations. It could be used for disencumbering processors. They are getting much faster and the performance of our systems could get much higher with this disencumbering, increasing the users' experience.

My job was to study snake algorithms and choose one to implement it on a GPU. While this complete algorithm runs, it has to use as few CPU time as it can.

I chose this algorithm because it is very slow and it have a big load. It is used all over the world for example in cartography, medical imaging and security, moreover at airports for face recognition.

There are lots of implementations of it but I chose the snake algorithm which uses gradient vector flow, by using mostly the GPU and least the CPU.

In my thesis I show how far I got, what the main particles of the operation of the algorithm are, how I realized it, what advantages and disadvantages it has. Furthermore I show what kind of softwares it is possible to use for writing programs with hardware acceleration and when it is worth taking of their advantages.

Downloads

Please sign in to download the files of this thesis.