基于小波变换和奇异值分解的剪纸纹样识别  

Paper cut-out patterns recognition based on wavelet transform and SVD

作　　者：张显全[1] 李国祥[1] 秦芳远[1] 

机构地区：[1]广西师范大学计算机科学系,广西桂林541004

出　　处：《计算机工程与设计》2010年第10期2280-2282,2286,共4页Computer Engineering and Design

基　　金：广西自然科学基金项目(0447035);广西自然科学基金项目(0832104);重庆市教委研究基金项目(KJ081309);广西教育厅基金项目(200607MS135);广西研究生教育创新计划基金项目(2008106020812M259)

摘　　要：针对民间传统剪纸艺术的计算机创作问题,在分析剪纸艺术特点的基础上,提出一种基于小波变换和奇异值分解的剪纸纹样识别方法。首先对剪纸纹样图像进行归一化和二值化处理,然后应用小波变换提取剪纸纹样图像的低频分量并进行奇异值分解,最后通过对奇异值进行归一化和降维处理作为最终的特征向量,利用最近邻分类器进行模式识别。实验结果表明,该方法能够有效地去除噪声干扰,较好的识别有一定艺术夸张变形的剪纸纹样。Aiming at the problems of the traditional paper cut-out art for computer creation,after analyzing the characteristics of paper cut-out art,a method of paper cut-out patterns recognition based on wavelet transform and singular value decomposition is proposed.The paper cut-out image normalized and binaryzation are discomposed in low frequency component by tow-dimension discrete wavelet,then distilled the singular value from low frequency component,after reducing dimension processing and making unitary dealing the characteristic vector of the image is obtained.The eigenvector is applied to nearest neighbor classifier to recognition.Experiments show that this method are efftive in diminishing the disturbing of the noise and recognizing paper cut-cut patterns of exaggerative deformations.

关 键 词：小波变换 低频分量 奇异值分解 特征提取 纹样识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]