误差可控的细分曲面图像矢量化  

Towards Error-controllable Image Vectorization Based on Subdivision Surfaces

作　　者：陈爱芬 李桂清[1] 王宇攀 聂勇伟 Chen Aifen;Li Guiqing;Wang Yupan;Nie Yongwei(School of Computer Science and Engineering, South China University of Technology, Guangzhou 510640)

机构地区：[1]华南理工大学计算机科学与工程学院,广州510640

出　　处：《计算机辅助设计与图形学学报》2017年第12期2197-2203,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61572202;61602183);广东省自然科学基金重点项目(S2013020012795);广州市科技计划(201707010140);华南理工大学中央高校培育项目

摘　　要：为了提高矢量化图像的重构质量,提出一种基于细分曲面的误差可控矢量化算法.首先提取图像特征,构建特征约束的初始网格,并利用二次误差度量方法简化初始网格,得到特征保持的基网格;然后利用带尖锐特征的Loop细分曲面拟合图像颜色,得到控制网格;最后计算重构图像的误差,对控制网格进行自适应细分,直至重构误差达到用户需求.实验结果表明,该算法能够大幅度提高初始重构结果的质量,并在一定程度上做到误差可控.To improve the reconstruction quality of vectorization images,an error-controllable vectorization algorithm based on subdivision surfaces is proposed.Given an image to be vectorized,it first extracts feature lines,constructs a dense initial mesh,and simplifies the initial mesh using the quadric error metric algorithm to obtain a base mesh.A Loop subdivision surface with sharp feature settings is then employed to fit the color height fields of the image.Finally,it adaptively inserts new control points into the control mesh for reducing the approximating error.Experimental results show that the proposed algorithm can obviously improve the quality of the reconstructed images compared to the original fitting results,and achieve error-controllable fitting to some extent.

关 键 词：图像矢量化 图像特征 网格简化 细分曲面 误差可控 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]