【摘要】 共生矩阵是一种常用的纹理特征提取方法,针对其易受噪声和图像中实体方向变化的影响等问题,提出一种新型共生矩阵纹理特征提取方法.该方法将傅里叶变换应用于原始图像,利用频域信息确定显著方向集合,把共生矩阵方法应用于显著方向集合以提取纹理特征.在进行图像匹配时,先在显著方向集合间进行相似度计算,后在纹理特征向量间进行相似度计算,显著方向集合中元素的显著性作为权重被引入纹理特征向量间的计算.实验结果表明,与传统共生矩阵方法相比,该方法针对有向纹理和无向纹理均具有较好的性能.更多还原

【Abstract】 Co-occurrence matrix is one of commonly used approaches for texture feature extraction.A novel approach to extract texture features using co-occurrence matrix was presented in order to reduce the effect of noise and entity direction change.Fourier transform was used for original image,and the frequency domain information was used to determine the dominant direction set.Co-occurrence matrix was used for the dominant direction set to extract texture features.At matching stage of image,the similarity calculation was carried out between the dominant direction sets.Then the similarity calculation was performed between the texture feature vectors,where the dominance of elements in the sets was introduced as weight.The experimental results show that the proposed approach has better performance for both directional textures and non-directional textures than co-occurrence matrix.更多还原