Efficient and robust 3D line drawings using difference-of-Gaussian

Zhang, Long; Xia, Jiazhi; Xiang, Ying; He, Ying; Müeller-Wittig, Wolfgang; Seah, Hock Soon

doi:10.1016/j.gmod.2012.03.006

Cited by 11 publications

(8 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Inspired by their use in image processing [MH80], Zhang et al . [ZXY*12] adapted the difference of Gaussians (DoG) concept to surface meshes for the depiction of characteristic lines. Their main idea is to apply two different Gaussian kernels

G_{σ_{e}}, G_{σ_{r}}

to the illumination

f : = ⟨ n, l ⟩

of the surface.…”

Section: Low‐level Illustrative Visualization Techniquesmentioning

confidence: 99%

A Survey of Surface‐Based Illustrative Rendering for Visualization

Lawonn

Viola

Preim

et al. 2018

Computer Graphics Forum

View full text Add to dashboard Cite

In this paper, we survey illustrative rendering techniques for 3D surface models. We first discuss the field of illustrative visualization in general and provide a new definition for this sub‐area of visualization. For the remainder of the survey, we then focus on surface‐based models. We start by briefly summarizing the differential geometry fundamental to many approaches and discuss additional general requirements for the underlying models and the methods' implementations. We then provide an overview of low‐level illustrative rendering techniques including sparse lines, stippling and hatching, and illustrative shading, connecting each of them to practical examples of visualization applications. We also mention evaluation approaches and list various application fields, before we close with a discussion of the state of the art and future work.

show abstract

G_{σ_{e}}, G_{σ_{r}}

to the illumination

f : = ⟨ n, l ⟩

of the surface.…”

Section: Low‐level Illustrative Visualization Techniquesmentioning

confidence: 99%

A Survey of Surface‐Based Illustrative Rendering for Visualization

Lawonn

Viola

Preim

et al. 2018

Computer Graphics Forum

View full text Add to dashboard Cite

show abstract

“…The abstracted shading method (Lee et al, 2007) demonstrated how these and lightingbased variants could be computed in image-space. Other variants based on image-space processing include Laplacian Lines (Zhang et al, 2009) and DoG lines (Zhang et al Zhang, Xia, Ying, He, Mueller-Wittig, and Seah (2012); Zhang, Sun, and He (2014)). In addition to speed, image-space lines have the advantage that they automatically remove clutter as a function of image-space line density, although, like all image-based methods, they potentially lose some fine-scale precision and control.…”

Section: Survey Of Feature Curvesmentioning

confidence: 99%

Line Drawings from 3D Models: A Tutorial

Bénard¹,

Hertzmann²

2019

View full text Add to dashboard Cite

This tutorial describes the geometry and algorithms for generating line drawings from 3D models, focusing on occluding contours.The geometry of occluding contours on meshes and on smooth surfaces is described in detail, together with algorithms for extracting contours, computing their visibility, and creating stylized renderings and animations. Exact methods and hardware-accelerated fast methods are both described, and the trade-offs between different methods are discussed. The tutorial brings together and organizes material that, at present, is scattered throughout the literature. It also includes some novel explanations, and implementation tips.We believe that these topics ought to be known by anyone interested in understanding the curves in visual representational art. It is one where computer graphics can make a unique contribution. Arguably, the algorithmic simulation of line drawing is a crucial step in understanding visual art. Occluding contoursThis tutorial focuses on the curves known as occluding contours or, simply contours. In some computer graphics research, these have been called silhouettes, though the silhouettes are technically a separate set of curves, as we will describe below.The occluding contours of a simple 3D object are shown in Figure 1.1. As a first definition, suppose we have a 3D object that we wish to render from a specific viewpoint. The occluding contours are surface curves that separate visible parts from invisible parts. By rendering the visible portions of these 3D curves together with the object, we get a basic line drawing (Figure 1.1b).There are many different contour detection and rendering algorithms, and some significant tradeoffs between them. The most important tradeoff is between simple algorithms that produce approximate results, and more complex algorithms that give more precision, control, and stylization capabilities. Just rendering reasonable-looking contours as solid black lines is very straightforward for a graphics programmer. These most basic algorithms can be implemented in a few additional lines of code in an existing renderer, and have been implemented in many real-time applications, including many popular video games (one of the earliest was Jet Set Radio in 2000). However, if we wish to stylize the curves, for example, by rendering the curves with sketchy or calligraphic strokes (Figure 1.1d), things become more difficult. With a bit of perseverance, renderings with distinctive and lovely styles can be created. At times, these renderings may still contain topological artifacts that are not suitable for very high-end production. High-quality algorithms that remove these artifacts are more complex; in the extreme, no provably-correct algorithm exists for this problem. However, there are a number of partial solutions that are good enough to be used in many circumstances, and we discuss in detail the issues involved.Note that, formally, the occluding contour and occluding contour generator are separate curves in 2D and 3D. However, we will frequently use the te...

show abstract

“…Vergne et al [2008] introduced the apparent relief, which extracts convexity information in the object-space and curvedness information from normal variations in the image-space, and combines both sources in a single shape descriptor. Zhang et al [2012] generalized the difference-of-Gaussian (DoG) edge detector to triangle meshes. In the object space, their algorithm smoothes the vertex normal twice with different Gaussian kernel sizes.…”

Section: Previous Workmentioning

confidence: 99%

“…This illumination model has proven effective for extracting feature in PELs [Xie et al 2007], Laplacian lines [Zhang et al 2009] and DoG lines [Zhang et al 2012]. τ = 0.980 τ = 0.990 τ = 0.995 Figure 3: The parameter τ is like a thresholding which controls the number of splatting lines.…”

Section: Illumination Splattingmentioning

confidence: 99%

“…Our approach is based on two widely-studied techniques: the Difference-of-Gaussian (DoG) for edge detection and the splatting for rendering point clouds. The DoG edge detector has proven to be effective and efficient in terms of generating high-quality line drawings from static images [Kang et al 2007], videos [Kyprianidis and Kang 2011], and 3D models [Zhang et al 2012]. Given a 2D signal (such as an image) as input, the DoG operator extracts the high-frequency feature lines by first filtering the signal twice, using Gaussian filters with different kernel sizes, and then subtracting the resulting two filtered signals.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Splatting lines

Zhang

Sun

2014

Proceedings of the 18th Meeting of the ACM SIGGRAPH Symposium on Interactive 3D Graphics and Games

Self Cite

View full text Add to dashboard Cite

a) 1.2M points, 65 fps (b) 327K points, 180 fps (c) 512 × 512 × 134, 26 fps (d) 256 3 , 160 fps (e) 1.8M tris, 24fpsFigure 1: Our algorithm can generate visually-pleasing line drawings from various types of models in real time. AbstractWhile 3D line drawing techniques have become well established in the past decade, most are only suitable for polygonal meshes. This paper presents splatting lines, a unified framework for generating line drawings from various types of 3D models, including point clouds, volumes, and polygonal meshes. In contrast to the existing mesh-based approaches, our method takes a densely sampled point cloud as input. It renders two diffuse shading images using splatting with different parameters and then generates line drawings by subtracting these images. Our point-based approach can be easily extended to polygonal meshes and volumes by an efficient (iso-)surface sampling method. Our method is highly efficient and it does not require any pre-computation. It is also more robust to the mesh tessellation than existing mesh-based techniques. Experimental results indicate the advantages of our method in terms of both performance and quality.

show abstract

Efficient and robust 3D line drawings using difference-of-Gaussian

Cited by 11 publications

References 31 publications

A Survey of Surface‐Based Illustrative Rendering for Visualization

A Survey of Surface‐Based Illustrative Rendering for Visualization

Line Drawings from 3D Models: A Tutorial

Splatting lines

Contact Info

Product

Resources

About