Enhancing pencil drawing patterns via using semantic information

Pencil drawing is recognized as a typical non-photorealistic visual art form. It is attractive to automatically generate high-quality pencil drawings from real-world photographs. Traditional model-based methods are highlighted in their interpretability, but are limited in modeling scene semantics, leading to weak sketch lines or limited pencil tone rendering effects. To address this issue, we propose a novel pencil drawing generation method via explicitly leveraging image semantic information. Specifically, the sketch lines generated from the weak boundaries can be boosted by fusing the segmented boundaries. In addition, the segmentation mask provides the spatial guidance for diversifying the tone shading effects. By combining the enhanced pencil drawing patterns, the visual quality of the generated pencil drawings can be therefore strengthened. We conduct several experiments to validate the effectiveness of our method, including ablation studies and comparison with other methods, in which promising results can be obtained.