TITLE:
Discrete-Time Dynamic Image Segmentation Using Oscillators with Adaptive Coupling
AUTHORS:
Mio Kobayashi, Tetsuya Yoshinaga
KEYWORDS:
Discrete-Time Coupled Model, Dynamic Image Segmentation, Oscillatory Responses, One-Dimensional Oscillator
JOURNAL NAME:
International Journal of Modern Nonlinear Theory and Application,
Vol.5 No.2,
June
28,
2016
ABSTRACT: In this
study, we propose a novel discrete-time coupled model to generate oscillatory
responses via periodic points with a high periodic order. Our coupled system
comprises one-dimensional oscillators based on the Rulkov map and a single
globally coupled oscillator. Because the waveform of a one-dimensional
oscillator has sharply defined peaks, the coupled system can be applied to dynamic
image segmentation. Our proposed system iteratively transforms the coupling of
each oscillator based on an input value that corresponds to the pixel value of
an input image. This approach enables our system to segment image regions in
which pixel values gradually change with respect to a connected region. We conducted
a bifurcation analysis of a single oscillator and a three-coupled model.
Through simulations, we demonstrated that our system works well for gray-level
images with three isolated image regions.