Nonlinear Noise Suppression Edge Detection Scheme for Noisy Images

In classification and recognition of objects the most commonly used features are edges, which are the locations where the intensity values change more than a predefined threshold value. This paper presents an efficient nonlinear based edge detection algorithm for noisy images. The edges are obtained in single phase using nonlinear filters without regularization. The algorithm calculates gradients in both directions that is horizontal and vertical direction using a nonlinear filter. The slope in a noisy image will have impact on noise. In each case the difference of the forward and backward gradient is taken as the actual gradient to localize the actual edge pixel and to suppress the noise in a noisy image. This scheme, obtains efficient edge pixels from a noisy image without using regularization or increasing the computational complexity.