An extended flow-based difference-of-Gaussians method of line drawing for polyhedral image

This paper presents a non-photorealistic rendering technique for stylizing a photograph in the line drawing style. The flow-based difference-of-Gaussians (FDoG) filter constructs lines of style and quality, which usually used to extract the line drawing for some images. In this paper, we inspired by the polyhedral image feature (the line of which are mostly straight line), proposed the method based on the standard FDoG to extract the line drawing of polyhedral image. First, we detect line by fast line segment detector with a false detection control (LSD); then construct a smooth and coherent edge tangent flow (ETF) of the polyhedral images. On the basis of the ETF, the anisotropic FDOG is used to render the lines. This scheme gives clear line features for polyhedral images without much noise, which is also simple and easy to implement, and is robust. The experimental results prove that our algorithm is more efficient than the traditional FDOG approaches in render line drawing for polyhedral image.