Zapraszamy na Krakow Quantum Informatics Seminar organizowane wspólnie przez Katedrę Informatyki AGH i IBM Software Lab Kraków. Drugie spotkanie odbędzie się we wtorek 27.11.2018 w godzinach 9:00-10:00 w Centrum Informatyki w sali 1.20.
W programie:
Michał Krok, IBM Software Lab Kraków
Quantum walks in image segmentation
Abstract:
Random walk is a well-known computational framework that has found an application in various problems, mainly in simulations of some natural and social processes in physics, biology or economics. Based on the success of random walks there have been developed quantum random walks which, by utilization of extraordinary properties of quantum world, have a potential to overcome their classical counterparts [1, 2].
During the seminar I'd like to present the result of my research focused on elaboration of an algorithm for application of quantum walks to one of the elementary tasks of image analysis, namely image segmentation. Taking inspiration from Grady’s work [3] I propose three methods of image segmentation: two algorithms harnessing quantum walks and one, which is quantum walk inspired, but is not a full-fledged quantum solution. All three methods have been simulated on a classical computer and provided results of comparable accuracy to the reference Grady’s method [4].
The discussion will be preceded by an introduction of the basics of these two computational models: classical and quantum random walks.
References
[1] Y. Aharonov, L. Davidovich, and N. Zagury. Quantum random walks. Physical Review A, 48(2):1687–1690, 1993.
[2] Julia Kempe. Quantum random walks: An introductory overview. Contemporary Physics, 44(4):307–327, 2003.
[3] Leo Grady. Random walks for image segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 28(11):1768–1783, 2006.
[4] Michał Krok , Quantum walks in image segmentation; Master of Science Thesis supervised by Katarzyna Rycerz and Piotr Gawron , AGH University of Science and Technology, Department of Computer Science, Krakow, Poland, 2018, http://dice.cyfronet.pl/publications/filters/filter_MSc_Theses