OCICETE 2017 Abstracts


Short Papers
Paper Nr: 1
Title:

How the GDPR will Affect the System Design? Four Main Novelties

Authors:

Gizem Gültekin Varkonyi

Abstract: Personal data protection now reaches its top attention with the developments in the European legislation. New General Data Protection Regulation of the European Union will enter into force in 2018, but from the day that it was announced, it was clear that it would bring new administrative and technical measurements to the corporates. In this paper, we will present the technical novelties that the GDPR brings for the corporates who are offering their services through systems or software.

Paper Nr: 2
Title:

Novel Approach of Reaction-diffusion Network for Image Processing and Computer Vision

Authors:

Atsushi Nomura, Koichi Okada and Yoshiki Mizukami

Abstract: Diffusion processes are widely found in nature. In a chemical reaction system, a reaction occurs at a particular point of its chemical solution, and its chemical substances diffuse, simultaneously. A reaction-diffusion system refers to a system of diffusion processes coupled with reactions. Such the reaction-diffusion system is generally described by a set of reaction-diffusion equations, which are time-evolving partial differential equations consisting of diffusion equations with reaction functions. The reaction-diffusion equations can be numerically solved by a finite difference method. We consider extending the reaction-diffusion system to a system of not diffusively but discretely coupled reaction elements. For example, we consider a network of a two-dimensional regular grid, and place a kind of excitable reaction elements at particular grid points. These elements are connected to each other in nearest four neighboring points. Thus, we call it 'reaction-diffusion network'. Each element has a state, which temporally changes according to external stimuli and its previous internal state. If the elements takes the uniform initial condition, all of the elements exhibit same behavior as time proceeds. Thus, nothing happens on the spatial distribution of the elements on the uniform initial condition. On the other hand, if the elements have the initial condition of a step-wise distribution consisting of high and low levels, only the elements located along the border of the two levels keep the excited level. Thus, the reaction-diffusion network performs an edge detection function of image processing. The author and his coworkers have proposed several algorithms for image processing and computer vision tasks, such as, edge detection and stereo disparity detection with the reaction-diffusion network. This presentation provides the novel approach of reaction-diffusion network, and its application to image edge detection. The algorithm using the approach is applied to datasets of real images, and the performance is quantitatively evaluated and compared with these of other typical edge detection algorithms.