上海大学学报(自然科学版) ›› 2010, Vol. 16 ›› Issue (1): 10-14.

• 通信与信息工程 • 上一篇    下一篇

拼接法纹理合成中的结构特征匹配

沈剑伟,王朔中   

  1. (上海大学 通信与信息工程学院,上海 200072)
  • 收稿日期:2008-07-30 出版日期:2010-02-28 发布日期:2010-02-28
  • 通讯作者: 王朔中(1943~),男,教授,博士生导师,博士,研究方向为信号处理、图像处理、信息安全等. E-mail:shuowang@shu.edu.cn
  • 基金资助:

    上海市重点学科建设资助项目(S30108)

Structural Matching in Texture Synthesis Based on Image Quilting

SHEN Jian-wei, WANG Shuo-zhong   

  1. (School of Communication and Information Engineering, Shanghai University, Shanghai 200072, China)
  • Received:2008-07-30 Online:2010-02-28 Published:2010-02-28

摘要:

 提出一种新的纹理合成技术,该方法对样本纹理图像进行采样,逐块合成纹理图像.在寻找最优候选匹配块时改变以往算法中仅匹配颜色相似度的做法,加入结构信息,提高了块边界结构的相似度.用该算法选择的最优匹配块更符合人的视觉特性,合成图像接缝区域的过渡更为流畅,结构单元的完整性明显优于原有方法.实验结果表明,该算法对结构性强的纹理具有良好的合成效果.

关键词: 纹理合成;相似度匹配;结构特征

Abstract:

Abstract: A new samplebased texture synthesis method is proposed in this paper. We sample an original small image containing textures and produce a synthesized texture image block by block using a modified image quilting technique. In a previous image quilting method, only color similarity is considered in seeking optimal matching candidate blocks. We propose to add structural information to the matching rule to seek better matched blocks. The obtained optimal blocks contain structural information leading to smoother transitions in the block boundary regions, therefore greatly improving perceptual quality. Experimental results show that the proposed method is particularly suited to generate synthesized texture from samples containing distinct structures.

Key words:  texture synthesis; similarity matching; structure characteristic

中图分类号: