Abstract: Conventional node-link based large graph visualization suffers heavy visual clutter due to edge crossing and node-edge overlapping, making the results difficult to explore.This paper proposes a novel method to visualize large-scale graph based on Line Integral Convolution.First, the connection relationship is clustered for each node with respect to the initial layout.The main connection orientations are then extracted to direct the generation of a synthetic vector field, which approximately describes the connection relationship in the input graph.Finally, LIC is employed to visualize this synthetic vector field.Experimental results demonstrate that our method is capable of releasing the visual clutter, and can reveal connection details hidden in edge crossing regions.