利用累加距离匹配函数的纹理规则度计算方法

摘要

规则度是纹理图像的主要特征之一，能够用于纹理图像的描述和分类。基于累加距离匹配函数(SDMF)提出一种纹理图像规则度计算方法，并建立规则-近似规则-随机纹理的分类模型和进行纹理粗细程度判定。首先针对纹理图像构建相应SDMF；然后利用二次求导和阈值划分提取显著峰谷；最后以峰值和谷值的数量，位置，关系为特征，构建周期规则度特征向量和纹理分类模型。以Brodatz图像库进行实验的结果表明，文中的规则度能够有效地区分规则图像、近似规则图像和随机图像，符合视觉感知，并与已有研究结果进行类比验证，具有更好的准确性；同时，由于SDMF的第一周期可以作为纹理基元的尺寸，因此可以扩展对纹理基元的粗细程度的分类。%Regularity is one of the main features of texture image, and it can be used for description and classification of texture images. A computation method of texture regularity based on summed-up distance matching function (SDMF) is proposed. On this basis, texture classification model for regular texture, ap-proximate regular texture and random texture is also proposed as well as texton roughness. First of all, SDMF of the texture image is constructed. Then, significant peaks and valleys of SDMF are extracted by second derivative and threshold division. At last, texture regularity computation and classification is charac-terized by the quantity, position and relationship among peaks and valleys. Experiment results based on Brodatz image database show that the proposed method can effectively distinguish different regular and ir-regular textures and has intuitive meaning. Compared with the existing method, the proposed method is more accurate. Meanwhile, the first periodicity of SDMF can be used as texture primitive size, that makes it as an extension of computation of degree of the texture roughness.