A simple and robust edge detection scheme based on coupled oscillatory diffusion system

Abstract

Edge is the most used and important segmentation feature in most of the object based image processing applications. Primary challenging issues with all the edge detectors are their adaptability for different scenes, noise immunity and most importantly complexity of implementation which can hinder real time performance for high resolution images. In this paper, we have proposed a novel, efficient and simple edge detection scheme which can be implemented on pixel level parallel architecture providing very high throughput. The detector is based on coupled oscillators applied over diffused image. Simultaneous solution of two time evolving linear difference equations produces edges on diffused intensity profile. The proposed system is inspired from FitzHugh Nagumo oscillator based edge detector but comparatively much simpler and produces better output on the standard precision-recall benchmark. The proposed detector is less dependent on control parameter (threshold) variation and thus suitable for real time edge detection of live video.