首页 | 本学科首页   官方微博 | 高级检索  
     

一种新的纹理合成算法
引用本文:郭小明.一种新的纹理合成算法[J].辽宁石油化工大学学报,2010,30(1):76-80,95.
作者姓名:郭小明
作者单位:辽宁石油化工大学理学院,辽宁抚顺,113001
基金项目:辽宁省教育厅高校科研项目,辽宁石油化工大学重点学科建设资助项目 
摘    要:现有纹理合成算法并不能保证合成物体的完整性,也不能使其分布与样本纹理保持一致。提出了一种分层纹理合成算法。首先,采用归一化图像分割算法提取样本纹理中的视觉显著性前景物体,继而采用图像修复算法修复背景,然后统计前景物体在样本纹理中的分布共生矩阵,以此作为约束,在背景图像上确定前景物体的分布位置。实验结果表明,不仅可以对前景物体独立进行各种变换,得到任意分布的合成纹理,而且可以保证前景物体的完整性。

关 键 词:纹理合成  图像分割  分布

A Novel Layered Texture Synthesis Algorithm
GUO Xiao-ming.A Novel Layered Texture Synthesis Algorithm[J].Journal of Liaoning University of Petroleum & Chemical Technology,2010,30(1):76-80,95.
Authors:GUO Xiao-ming
Abstract:Current texture synthesis algorithms can neither preserve the integrity of the salient object,nor keep the same distribution of the salient object in the synthesized textures as that in the source texture.A novel layered texture synthesis algorithm was proposed.Firstly,the salient objects were extracted by adopting the normalized cut image segmentation method.Then an image imprinting algorithm was applied to fill the holes caused by salient object extraction in the background to generate a completed background sample,with which we can synthesis the full background image of the target texture.The experimental result shows that this algorithm can not only get the distribution of the salient object in the synthesized textures but also preserve the integrity of the salient object.
Keywords:Texture synthesis  Image segmentation  Distribution
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号