This paper presents an edge-enhanced isotropic nonlinear filter to extract strokes from cartoon images. The algorithm starts by removing most non-stroke regions that have high brightness values using an adaptive threshold. According to the property of dark outline in cartoon images, A Laplacian of Gaussian filter is applied to enhance the isotropic responses of boundary pixels in the stroke. Compared to existing stroke extraction methods, our algorithm can extract strokes in its entirety, and faithfully preserve the details on the stroke boundaries, especially for noisy images. Experimental results show that the proposed algorithm can produce more robust and accurate extraction results on a variety style of cartoon images.