Perceptually-based Stroke Pattern Synthesis

Abstract

We present a novel method to synthesize stroke patterns from a user-specified vector reference pattern. The main idea is inspired by research in visual perception. We first extract the features which have an important impact on visual perception from the reference pattern, and then adopt a three-stage approach to synthesize new patterns. The synthesis starts with a randomly generated pattern whose element density is the same as that of the reference pattern, and then we substitutes the elements iteratively to optimize the proportion and neighborhood features. Finally, we iteratively adjust the elements' position to ensure that the distance feature of the target pattern is close to that of the reference one. The method can synthesize most of the stroke patterns with the element distribution from uniform to non-uniform, and from overlapping to non-overlapping