中文摘要：

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。

英文摘要：

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.