双树复小波域基元共生矩阵的纹理特征提取方法

Texture Feature Extraction Algorithm Based on Primitive Co-occurrence Matrix in Dual-tree Complex Wavelet Domain

提出了一种双树复小波域基元共生矩阵的纹理特征提取方法。利用双树复小波域作为分析域,以纹理基元作为依据,来描述纹理的局部特征,提取图像纹理特征值。实验结果表明,该方法能有效提取出多尺度、多方向的纹理特征,并兼顾纹理局部随机性和整体规律性,所提取的纹理特征具有良好的聚类分离度和类内样本差异性。

This paper presents an extraction method of dual-tree complex wavelet domain primitive co-occurrence matrix.Local features of texture are described and characteristic values of image texture are extracted by using dual-tree complex wavelet domain as the analysis domain,and texture primitive as the basis.Experiments show that multi-scaled and multi-directional texture features can be effectively extracted by using this method while paying attention to local randomness and overall regularity.Thus the extracted features have good clustering separation and within-class sample difference.