看完了论文《Over Coat: AnImplicit Canvas for 3D Painting》给我的感觉就是即便是在siggraph这样的国际一流大会上，论文运用的数学公式其实也是很简单的。
一方面是所谓的大道至简。很多时候，最好的解决方法往往是很简单的方案。
另一方面，也许老外的数学真不怎么样。这篇文章倒是给我建立了一些信心，自己不再像研一的时候觉得老外论文的数学很深奥了。真的就那么回事。
论文阅读笔记如下，部分名词和较简单的语句没有翻译，自己重新寽了一下写作时不甚明晰的逻辑。
文件已经上传到新浪共享，可以搜索“Over Coat: AnImplicit Canvas for3D Painting”设置了1分的金币，目的是想知道大概多少人下载过，望理解。
论文阅读《Over Coat: AnImplicit Canvas for3D Painting》
0、摘要
主要成果：present a technique togeneralize the 2D painting metaphor to 3D that allows the artist to treat thefull 3D space as a canvas
需要达到的要求：Strokes painted in the2D viewport window must be embedded in 3D space in a way that gives creativefreedom to the artist while maintaining an acceptable level of controllability.
解决方案：
l canvas：concept defined implicitly by a 3Dscalar field
l artist：shapes the implicit canvas bycreating approximate 3D proxy geometry
l Anoptimization procedure：embed painted strokesin space by satisfying different objective criteria defined on the scalar field.àallows us to implement tools forpainting along level set surfaces or across different level sets.
l aunified painting/sculpting metaphor: fine-tuning the implicit canvas
l Asculpting tool:paint into the implicit canvas, this tool creates a local change in the scalarfield that results in outward or inward protrusions along the field’s gradientdirection.
l 3Dstroke rendering:visibility ambiguity(inherent) with a depth offsetting method(suited forhardware acceleration)
1、背景
文章提到NPR的近些年的发展，并着重提到了几个有代表性的成果：
1、水彩合成【Bousseau et al. 2006】
2、毛发、草、树的风格化显示【Kowalski et al. 1999】
3、显要边界【Judd et al. 2007】
4、实时的非真实感渲染【Markosian et al. 1997；Praun etal.2001】
作者总结，这些技术的共同之处即它们都是实现3d到2d映射的一种算法，不同的只是映射后表现的形式不同。
技术发展路线：
1、Meier’spainterly rendering system。在众多的NPR研究成果之中Meier’s painterly renderingsystem[1996]是与作者目标非常相关的方法，这个论文是他们研究的开端。文章涉及的核心思想：Particles attached toan object’s primary-space surface are used to render secondary-space brushstrokes so that the strokes stick to the object as the camera moves.
à
2、DeepCanvas。上面这篇文章中包含的看似简单的思想引导了Disney的DeepCanvas技术【Katanics and Lappas2003】。这篇文章的创新之处在于：replaces Meier’sprocedurally generated particles with an artist-driven painting system. Paintedstrokes are projected on the object’s surface and stored along with all dataassociated with the painting system. A new view is rendered by “repainting,” orplaying back all recorded painting operations, using the camera’s new viewtransformation.
与Deep Canvas同时期还有学者提出了一个关注于交互性的绘画相关概念Teece’stechnique
à
3、WYSIWYGNPR。The WYSIWYGNPR system of Kalnins and colleagues【2002】又在这条研究道路上向前走了一步。他们的贡献是解决了：how algorithmicrendering techniques such as silhouette stylization or hatching can becontrolled directly by the artist via a painting interface.
à
4、PaintEffects。Maya PaintEffects [Paint Effects 2011]projects painted strokes onto the surface of scene geometry and uses them asseed points to create new geometric primitives such as grass or flowers.
|
综上这些技术 |
共同点 |
这样做的缺点 |
本文的方法 |
|
Meier’s painterly rendering system |
embed strokes directly on the surface of an object or along object features such as silhouettes |
the strokes must conform to the surfaces of their associated primary-space objects. à the functionality of these systems is intimately tied to and restricted by traditional surface representations such as polygon meshes and NURBS surfaces. àThe exacting nature of these surface representations may be at odds with an artist’s particular style, hindering their ability to realize their artistic vision. |
built an artist-driven 3D painting system that employs a stroke-based renderer to generate temporally coherent expressive imagery. allows strokes to be embedded anywhere in space |
|
Deep Canvas |
|||
|
Teece’s technique |
|||
|
WYSIWYG NPR |
|||
|
Paint Effects |
文章提出的这种方法可以达到paint anywhere inspace，但同时也在可用性和控制上遇到了前人工作未曾涉及的挑战。作者利用implicit canvas和optimization formulation解决了上述问题。
部分研究者已经实现利用特定的输入设备达到在三维空间中布放颜色的难题：
1、The Cave-Painting system of Keefeand colleagues： [Keefe et al. 2001] uses motion capture in avirtual-reality cave to allow the artist to directly author scenes composed ofribbons, tubes, and other primitives using hand gestures.
2、Schkolne and colleagues’ SurfaceDrawing work：2001. enables organicshapes to be modelled using hand gestures in a semi-immersive VR setup.
3、Keefe and colleagues [2007]：解决了基于手势系统中经常遇到的可控性的问题。他们propose a method that uses ahaptic device and 6-DOF trackers to draw lines in space in a more precisefashion。
文章在如下两个方面超越了前人：
一、OverCoat extends the tablet-based2D digital painting metaphor to 3D without the need of special hardware, thusmaking it more accessible and familiar to work with for artists.
二、OverCoat’s embeddingprocedure, brush model, and rendering algorithm merge 2D and 3D concepts to enablepaintings that retain their 2D expressiveness when viewed from any angle.
一些商用建模软件如maya，mudbox，zbrush提供了创建3D proxy geometry的工具，这帮助我们完成了部分工作。
一些基于手绘的建模工具【Igarashi et al. 1999;Nealen et al. 2007】尤其是那些使用隐式表面为表现形式的如【Karpenko et al. 2002;Schmidt et al. 2005; Bernhardt et al.2008】都很适合于本项目因为the proxygeometry need only provide a rough guide to control the implicitcanvas但是implicitsurface representation从来不是直接被渲染的。
其他一些相关工作【Cohen et al. 2000;Bourguignon et al. 2001; Tolba et al. 2001; Rivers et al. 2010】探索了一些有关3D绘画的有趣的概念，但是并未细致的介绍3D painting，并且cannot easily accommodate theexpressive aesthetic we target with OverCoat。
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
详情在新浪爱问共享上搜索“Over Coat: AnImplicit Canvas for3D Painting”