Edge Detection Based on Gradient Histogram and Variance Within Clusters

In this paper,a new method for edge detection from complex scenes based on gradient histogram and variance within clusters is proposed(cluster algorithm based on gradient histogram,CAGH).After analysing gradient histogramming by "Non-maximal suppression",edges can be extracted with the gradient threshold based on the variance within clusters.In comparision with other methods for edge detection of vehicle license plates from complex scenes,this method has better adaptability and more efficiency,which generates one pixel width edges with good connectivity and independence,is good to subsequent feature extraction and pattern recognition.