Simple gray-scale image segmentation method based on locally coupled neural oscillator network

Abstract

The invention discloses a simple gray-scale image segmentation method based on a locally coupled neural oscillator network, comprising the steps of establishing a visual image input layer, establishing a neuron oscillator network oscillatory layer and establishing an inter-objective separation layer, wherein The visual image input layer is in charge of establishing one-to-one correspondence between the visual images and the neuron oscillators in the neural network on the neuron oscillator network oscillatory layer; the neuron oscillator network oscillatory layer is in charge of establishing a dynamical system model for each neuron oscillator by simulating the function of a functional column in the visual cortex of the human brain in processing the visual images and ensuring each oscillator to generate oscillation and generate synchronous oscillation under the local coupling effect of adjacent oscillators; and the inter-objective separation layer is in charge of realizing separation of objective regions on the visual images by adopting a desynchronization mechanism according to the synchronous oscillation result. In the method, the parameter setting requirements for the neuron oscillators to generate oscillation and synchronous oscillation are analyzed, and the method plays a guiding role and has theoretical significance in knowing and understanding image segmentation.

Description

A Simple Grayscale Image Segmentation Method Based on Locally Coupled Neural Oscillator Networks technical field The invention relates to the technical field of image processing and neuron network, in particular to a system realized in the form of a locally coupled neuron oscillator network for processing the separation of different target areas under the same background in a visual image. Background technique Human eyes can easily perceive and understand visual images in a short period of time, but researchers still know little about the mechanism. However, a large number of physiological experiments have shown that the visual organs may first extract various local features of the object and transmit them to the central nerve in parallel. The problem is how to reintegrate these features in the central nervous system. Separation of different objects (regions of interest) falls into this category of problems. To deal with this, Hebb theorized that in order to detect an obj...

AI Technical Summary

Problems solved by technology

Human eyes can easily perceive and understand visual images in a short period of time, but researchers still know little about the mechanism

However, a large number of physiological experiments have shown that the visual organs may first extract various local features of the object and transmit them to the central nerve in parallel. The problem is how to reintegrate these features in the central nervous system. Separation of different objects (regions of interest) falls into this category of problems

To deal with this problem, Hebb proposed a theory that in order to detect an object, a group of neurons needs to increase their average firing rate, but this leads to a new problem - the superposition problem

Images