Quad-pixel edge detection using neural network

One of the most fundamental features of digital image and the basic steps inimage processing, analysis, pattern recognition and computer vision is the edge ofan image where the preciseness and reliability of its results will affect directly onthe comprehension machine system made objective world. Several edge detectorshave been developed in the past decades, although no single edge detectors havebeen developed satisfactorily enough for all application. In this paper, a new edgedetection technique was proposed based on the BP neural network. Here, the edgepatterns of a quad-pixel in binary images were classified into 16 possible types ofvisual patterns. In the following, after training the pre-defined edge patterns, theBP neural network was applied to correspond any type of edges with their relatedvisual patterns. Compared with traditional edge detection techniques, the resultsdemonstrate that the new proposed technique, improved the computations mass andmathematical complexity, turns out better.

Keywords:

Edge detection , Image binarization , Image processing , Neural Networks , Threshold