Based on the concept of discrete singular convolution (DSC), Hou and Wei (2002) introduced a novel edge detection method using one of singular kernels-the delta Shannon. So called the DSC anti-noise edge detector (DSCANED), this method is capable of extracting edges against noise. In this study, we further introduce another two kinds of kernels, delta Dirichlet and de la Vallée Poussin to construct the Dirichletbased and the Poussin-based DSCANEDs which then are compared with the Shannon-based one. The salt and pepper noise of different densities is added to a set of images as well as a standard binarized circular pattern for generating several noisy test samples. Experiments indicate that the performance of Dirichlet-based DSCANED is outperformed. It is speculated that such kernel has one more parameter which can be optimized to achieve better results.