结构化矢量图的模式样本合成与操控

项目名称： 结构化矢量图的模式样本合成与操控

项目编号： No.61303147

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 周世哲

作者单位： 湖南大学

项目金额： 28万元

中文摘要： 本研究项目将探索矢量图形的模式感知合成技术，以对其进行缩放和操控。许多矢量图中常常包含按一定模式排列的离散子图, 同时这些子图往往由规整的外框所包围。给定这样一个小尺寸的模式样本，艺术家在进行建筑制图等任务时往往会自然地将样本中的子图在目标尺寸上进行拼贴重排以维持这种模式。该过程耗时且容易出错。所以，我们将结合模式感知形变方法和基于样本的合成技术来解决矢量图的合成问题。现有模式感知形变法单独处理每个模式子图并忽略了封闭子图带有的区域信息，这导致已知方法无法输出带有正确拓扑的完整的合成结果。我们将研究如何在合成过程中保存这些区域信息，从而使得我们的方法能处理欠规则的或者带有任意拓扑结构的输入矢量图。为了引导合成，我们将首先设计一套自动的矢量形状分析流程，该流程能分析出如按曲线排布的各类模式。分析所得信息将用于引导在网状结构上执行多方向模式合成。我们开发的系统可使得用户轻松制作高质量的矢量图。

中文关键词： topology descriptor；vector texture；pattern-aware retargeting；graphics content creation；dynamic programming

英文摘要： This research project will explore a novel pattern-aware synthesis technique for resizing and direct manipulation of vector graphics. Many vector graphics contain regualar patterns consisting of multiple small elements bounded by structural frame shapes.Given such an exemplar,aritists often naturally try to manually preserve these patterns during content creation such as architectual ploting and planar design, which can be labor intensive, error-prone and requires artistic skills.We will explore the possiblities of applying the combination of pattern-aware deformation technique and by-example texture synthesis method for synthesizing such vectorial patterns automatically.Since existing pattern-aware deformation method treats shape elements separately and discard region information of closed elements,their outputs often lack compact representations and correct topologies. We plan to explore solutions to handle region information correctly and merge sliced elements in a global optimal order,which enable us to robustly handle vector graphics of which elements has various shape propoerties and aribitrary topologies.Before synthesis, we will develop an automatic computational framework that analyze pattern shapes including irregular distribution and rotational symmetry arrangement from the examplar. With prescribed s

英文关键词： 拓扑描述子；矢量纹理；模式感知重缩放；图形内容创建；动态规划