Image Segmentation Using Frequency Locking of Coupled Oscillators

TL;DR

This paper introduces a novel image segmentation method using a network of coupled oscillators that synchronizes to identify image regions, demonstrating improved accuracy and noise tolerance over traditional threshold-based algorithms.

Contribution

The paper presents a new oscillator-based system for image segmentation that leverages frequency locking, offering enhanced performance and noise robustness compared to existing methods.

Findings

Better segmentation accuracy than traditional methods

Higher noise tolerance in image processing

Effective on human face image dataset

Abstract

Synchronization of coupled oscillators is observed at multiple levels of neural systems, and has been shown to play an important function in visual perception. We propose a computing system based on locally coupled oscillator networks for image segmentation. The system can serve as the preprocessing front-end of an image processing pipeline where the common frequencies of clusters of oscillators reflect the segmentation results. To demonstrate the feasibility of our design, the system is simulated and tested on a human face image dataset and its performance is compared with traditional intensity threshold based algorithms. Our system shows both better performance and higher noise tolerance than traditional methods.