首页 | 本学科首页   官方微博 | 高级检索  
     

纹理的标准性和强标准性纹理的快速识别及合成
引用本文:古元亭,叶正麟,陈飞.纹理的标准性和强标准性纹理的快速识别及合成[J].计算机辅助设计与图形学学报,2005,17(4):712-718.
作者姓名:古元亭  叶正麟  陈飞
作者单位:西北工业大学数学与信息科学系,西安,710072;西北工业大学数学与信息科学系,西安,710072;西北工业大学数学与信息科学系,西安,710072
基金项目:国家自然科学基金(10071060)
摘    要:分析了纹理合成中时间开销最大的匹配过程,发现了对一类具有特征任意匹配性纹理,可以大大降低匹配精度甚至忽略匹配过程,从而加快纹理生成速度.通过对这类纹理特性的概括,引入了纹理标准性概念,利用可变尺度滑动窗口的灰度距离对标准性进行度量;并结合大量统计结果,给出纹理标准性区分标准,对强标准性纹理使用低精度的匹配或零搜索来合成纹理.实验结果证明,文中对于强标准性的纹理区分可靠有效,生成纹理的速度大大提高,且质量无明显降低.

关 键 词:纹理合成  纹理标准性  灰度距离  零搜索

Standard of Texture and Quick-Identification and Synthesis to Strict Texture
Gu Yuanting,Ye Zhenglin,Chen Fei.Standard of Texture and Quick-Identification and Synthesis to Strict Texture[J].Journal of Computer-Aided Design & Computer Graphics,2005,17(4):712-718.
Authors:Gu Yuanting  Ye Zhenglin  Chen Fei
Abstract:This paper first gives an analysis of the matching process in patch-based sampling texture synthesis, which takes most of the time in the whole process of the synthesis algorithm. We find that for a certain type of texture, with its texture features distributed uniformly, we can omit the matching process when they are synthesized, or we can make the match roughly. By summing up the properties with regard to the textures of this type, a concept of textures' strictness intensity is introduced that can be used as a measure of how much an image looks like a texture intuitively. Based on the principle, an algorithm is proposed to calculate the intensity coefficients of images according to the gray scale distance among slipping sub-windows in variable size and the calculation result helps to optimize the precision of matching process. Finally in the paper, we give the synthesis results of strict textures identified by our algorithm. The experiments show that in particular to the strict texture, the generating process may be improved, while the image quality is well reserved.
Keywords:texture synthesis  strictness intensity of texture  distance of gray scale  zero-search
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