Image Edge Detection by Savitzky-Golay Differentiator

Image edge referred as the abrupt changes in pixel intensity and it can offer the important features for recognizing an object. The conventional methods for detecting edges of an image involves convolution operation with the first-order derivative kernel such as Sobel and Prewitt mask. These masks have demerits of the high sensitivity to noise and limitation in defining size of kernel with only 3 x 3. This study proposes the two-dimensional extensions of one-dimensoional Savitzky-Golay first-order differentiator to detect image edges. The merits of the extended Savitzky-Golay differentiator are emphasized by comparing with the performance of Sobel and Prewitt edge operator in terms of the capability of noise suppression and the accuracy in locating edges.