In this paper, we present an order statistics based edge detector which has good edge detection characteristics in the presence of Gaussian and impulsive noise. The developed edge detector combines an order statistic
(OS) Laplacian operator which is...
In this paper, we present an order statistics based edge detector which has good edge detection characteristics in the presence of Gaussian and impulsive noise. The developed edge detector combines an order statistic
(OS) Laplacian operator which is designed to be robust with respect to noise, and a zero-crossing detector.
The edge detector computes the second derivative of the image using the OS Laplacian operator and the zero-crossing detection is followed to locate the edge points.OS Laplacian operation is carried out from the difference of the outputs of a median filter and a modified trimmed mean(MTM) filter.
The performance of the edge detector in the presence of Gaussian and impulsive noise is evaluated and compared to the performance of well established edge detectors such as a Laplacian of Gaussian(LoG) operator and a Sobel operator. Experimental results with both synthetic and real images show that the OS edge detector has the best edge detection characteristics in the presence of Gaussian and impulsive noise.