基于L邻域分割的结构性纹理合成方法  

Structural texture synthesis method based on L neighbor segmentation

作　　者：张雨禾[1] 耿国华[1] 刘伦椿[1] 周子骏 

机构地区：[1]西北大学信息科学与技术学院,陕西西安710127

出　　处：《哈尔滨工程大学学报》2014年第10期1265-1271,共7页Journal of Harbin Engineering University

基　　金：国家973计划前期研究专项资助项目(2011CB311801);国家自然科学面上基金资助项目(61172170)

摘　　要：针对纹理合成过程中L邻域整体匹配造成的误差,充分考虑了结构性纹理中像点之间的相关性信息,提出了一种新的L邻域匹配方法。取4个特殊的方向,对L邻域进行分割,利用纹理样图的灰度共生矩阵的相关值,对各个分割区域确定匹配优先级。采用基于优先级的多次匹配,找出L邻域的最佳匹配块。实验证明,该方法减少了传统的L邻域在匹配搜索过程中存在的计算复杂性和匹配误差,有效地提高了纹理合成中预处理的速度,在结构性纹理合成中获得了好的效果。This paper proposes a new L neighbor matching method for errors caused by overall L neighbor matching during the texture synthesis process with full consideration of the correlation information between pixels in structural texture. It segments the L neighbor along 4 special directions,marks each divided region with matching priority based on the Grey-level co-occurrence matrix（ GLCM） ＇ s IDM,and thereby determines the matching priority of each divided region. It also uses the multi-matching method based on priority to identify the best matching block of L neighbor. Experimental results proved that this method reduces the computational complexity and matching errors caused by traditional L neighbor matching. It effectively improves the preprocessing speed of texture synthesis and obtains good structured texture synthesis effect.

关 键 词：纹理合成 L邻域分割 灰度共生矩阵 结构性纹理 相关性信息 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]