Adaptive Blending of Multiple Network Layouts for Overlap-Free Labeling

Abstract

Conventional force-directed algorithms are known as a common approach to aesthetically drawing networks while they still suffer from self-overlaps especially when the network nodes are annotated with text labels. Incorporating space partitioning techniques including Voronoi tessellation are often effective to spare enough space around each node while this may incur different artifacts such as unexpectedly long edges and edge overlaps. This paper presents an approach to resolving overlaps among node labels by adaptively blending multiple layout forces applied to the respective network nodes. This is accomplished by extending our previous approach for transforming the force-directed layout into that obtained through the centroidal Voronoi tessellation. Our technical contribution lies in a novel algorithm for smoothing blending ratios associated with the network nodes so that we can adaptively explore the reasonable balance between the two layouts independently for each node. Experimental results will present that our new approach can produce well-balanced distribution of node labels while maximally avoiding the aforementioned unwanted visual artifacts.