基于同步振荡神经网络模型的图像分割

Image Segmentation Based on Synchronous Oscillation Neural Network Model

目前,国内外学者提出了很多图像分割算法,其中一些已经广泛应用于灰度图像和彩色图像分割。本文对图像分割方法进行了总结,并利用振荡神经网络模型的同步提出了对彩色图像分割的算法。我们在经典Kuramoto模型的基础上将全局耦合改为局部耦合,将相位耦合改为频率耦合并用它描述外界刺激引起的相位变化。通过引入瞬时频率,重建耦合神经元的活动曲线,将原彩色图像分别通过R、G、B三个通道,形成三个振动曲线,并将其进行叠加,根据一致同步性原理,得出一个新的彩色图像分割算法。

So far, many image segmentation algorithms have been proposed, and some of them are used widely for both gray image and color image segmentation. This paper presents a kind of color image segmentation algorithm based on harmonic which comes from superposition of simple harmonic waves. In order to describe the activity of neuron more appropriately, the original Kuramoto model is changed from phase coupling to frequency coupling and globally coupling is changed to locally coupling. The instantaneous frequency is introduced to represent the phase change as a consequence of external stimuli. The activity of the coupled neurons is reconstructed using instantaneous frequency. The pixel values of R, G, and B of color image are extracted and put into the network, three oscillating curves are produced, and they are superposed to produce the consonance, the new color image segmentation algorithm is formed according to the principle of synchronization of the consonance.