A Nonlinearly Weighted Sobel Edge Detector

Abstract

In this paper, a modified Sobel edge detector (SED) is proposed where the gradient is nonlinearly weighted. The modified SED is called nonlinearly weighted SED (NWSED). In the SED, only the difference of neighbor pixels is considered in the gradient calculation. However, in the NWSED the ratio of neighbor pixels is put into account. The NWSED is then analyzed and found of nonlinear weighting property. With the property, the difference of low gradients and high gradients is increased. To justify the proposed NWSED, simulations are performed. The results indicate that the NWSED has better performance than the SED in the preserving of edge information.