【作者】 闫怀平； 赵凤群； 戴芳； 张晓艳；

【Author】 YAN Huai-ping ZHAO Feng-qun DAI Fang ZHANG Xiao-yan School of Science, Xi’an University of Technology, Xi’an 710054, China

【机构】 西安理工大学理学院；

【摘要】 以往的纹理合成只是将两个匹配块之间对应像素的颜色差异作为评价相似性的度量,不能很好地把握样图的结构性信息,本文提出一种新的加入结构性信息的方法,用边缘跟踪算法与边界的链码表述方法相结合来记录样图的结构信息,并用这些结构信息来计算两个块之间的结构匹配误差。所提出的方法将结构差异作为相似性度量的重要因素之一,保证了图像间结构信息的连贯性,取得了比较理想的合成结果。更多还原

【Abstract】 Color difference between the corresponding pixels of the two texture patch is only viewed as the similarity measure in existing texture synthesis algorithm, so it is difficult to preserve the structure information of the sample. A new algorithm introducing structure information into texture synthesis is presented. This method integrate the edge tracking algorithm with chain code method to record the structure information of sample ,and use the structure information of sample to calculate the structure matching error. In this paper, the structure difference is viewed as the important similarity measure to ensure the structure continuity within a texture, which gets satisfactory results.更多还原