A fast local laplacian filter for high quality suggestive contours generation

Line drawings play an important role in shape description due to they can convey meaningful information by outstanding the key component and distracting details or ignoring less important. Suggestive contours are a type of lines to produce high quality line drawings. To generate those contours, we can generally start from two aspects: from image space or object space. The image space strategies can not only extract suggestive contours much faster than object space methods, but also don't require the information of the 3D objects. However they are sensitive to small noise, which is ubiquitous in the digital image. In this paper, before extracting lines, we apply an accelerated structure-preserving local Laplacian filter to smooth the shaded image. Through our experiments, we draw the conclusion that our method can effectively suppress the redundant details, generating a cleaner, higher quality line drawing by image space methods, and can compare with the result by object space ones.